JP2023017962A - Hdl-associated protein biomarker panel detection - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for accurately determining the risk of cardiovascular disease (CVD) for a subject with, or suspected of having, CVD or other HDL related disease.

SOLUTION: The present invention provides methods, systems, and compositions for detecting one or more HDL-associated proteins (e.g., ApoC3; ApoC3 and ApoA1; ApoC3 and SAA1/2; or proteins in Biomarker Panels 1-30) in a sample from a subject with, or suspected of having, cardiovascular disease (CVD) or other HDL related disease. In certain embodiments, such methods, systems, and compositions are used to determine the approximate risk of CVD (or other disease) for a subject, and/or the approximate cholesterol efflux capacity (CEC) of a sample.

SELECTED DRAWING: Figure 1

COPYRIGHT: (C)2023,JPO&INPIT

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to US Provisional Application No. 62/365,175, filed July 21, 2016, which is hereby incorporated by reference in its entirety.

FIELD OF THE INVENTION Herein, one or more HDL-related proteins (e.g., ApoC3; ApoC3; and ApoA1; ApoC3 and SAA1/2; or proteins of biomarker panels 1-30). In certain embodiments, such methods, systems, and compositions are used to determine the approximate risk of CVD (or other disease) in a subject and/or the approximate cholesterol efflux capacity (CEC) of a sample. do. In certain embodiments, the systems and compositions comprise a sample from a subject having or suspected of having CVD and an HDL-related binding agent or mass spectrometry standard.

Despite recent advances in both our understanding of the pathophysiology of cardiovascular disease and techniques for imaging atherosclerotic plaques, it remains difficult to accurately determine risk in patients with stable heart disease. be. Clinically unidentified high-risk patients who have not undergone aggressive risk factor modification and who experience significant cardiac adverse events are a significant problem. Similarly, refocusing finite health care resources to those most likely to benefit requires more accurate identification of low-risk subjects. Current clinical risk assessment tools, mostly related to algorithms developed from epidemiologically based studies of untreated primary prevention populations, are applicable in the setting of high-risk and medicated cardiac outpatients. Limited. Despite the high interest, efforts to incorporate more comprehensive array-based phenotyping techniques (e.g., genome arrays, proteome arrays, metabolome arrays, expression arrays) to improve cardiac risk stratification are nascent; We have yet to reach an efficient and robust platform that can meet the high speed demands of clinical practice.

As used herein, one or more HDL-related proteins (e.g., ApoC3; ApoC3 and ApoA1; ApoC3 and SAA1/2; or proteins of biomarker panels 1-30) are provided. In certain embodiments, such methods, systems, and compositions are used to determine the estimated risk of CVD (or other disease) in a subject and/or the estimated cholesterol efflux capacity (CEC) of a sample. do. In certain embodiments, the systems and compositions comprise a sample from a subject having or suspected of having CVD and an HDL-related binding agent or mass spectrometry standard.

In some embodiments, the methods provided herein comprise detecting the level of at least one HDL-related protein in a sample from the subject, wherein the at least one HDL-related protein contains ApoC3 and the subject has or is suspected of having cardiovascular disease. In certain embodiments, the sample is a purified high-density lipoprotein sample. In other embodiments, the sample is selected from the group consisting of serum samples, plasma samples, and blood samples. In additional embodiments, the method normalizes the level of at least one HDL-associated protein with the estimated level of total HDL particles, or ApoA1, or HDL cholesterol to obtain a normalized value of the at least one HDL protein. Further comprising generating.

In certain embodiments, determining the approximate level of total HDL particles, or ApoA1, or HDL cholesterol in the sample comprises determining the level of an internal standard added to the sample, wherein the internal standard The material includes labeled HDL protein. In some embodiments, detecting the level of at least one HDL protein comprises adding a reagent to the sample, wherein the reagent digests the HDL protein in the sample (e.g., Lyc-C, pepsin, trypsin, etc.). In other embodiments, HDL-associated proteins are detected without prior digestion (eg, to detect intact HDL-associated proteins).

In other embodiments, detecting the level of at least one HDL-associated protein comprises performing an assay that detects ApoC3 or ApoC3 fragments on at least a portion of the sample. In certain embodiments, detecting the level of at least one HDL-associated protein further comprises adding a labeled or unlabeled ApoC3 protein or protein fragment standard to the sample and detecting the ApoC3 standard. include. In a further embodiment, the ApoC3 calibrator comprises or consists of the amino acid sequence shown in SEQ ID NO:11. In certain embodiments, the assay is a mass spectrometric assay or an immunoassay.

In certain embodiments, the at least one HDL-associated protein further comprises serum amyloid A1 or 2 (SAA1 and/or SAA2, referred to herein as "SAA1/2"). In a further embodiment, detecting the level of at least one HDL protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments and ii) SAA1/2 or SAA1/2 fragments. In additional embodiments, detecting the level of at least one HDL-associated protein includes adding a labeled or unlabeled SAA1/2 protein or protein fragment standard to the sample and detecting the SAA1/2 standard. further including In certain embodiments, the SAA1/2 calibrator comprises or consists of the amino acid sequence set forth in SEQ ID NO:35.

In further embodiments, the at least one HDL-associated protein further comprises apolipoprotein 1 (ApoA1). In some embodiments, detecting the level of at least one HDL-associated protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments and ii) ApoA1 or ApoA1 fragments. In other embodiments, detecting the level of at least one HDL-associated protein comprises assays that detect i) ApoC3 or ApoC3 fragments, ii) SAA1/2 or SAA1/2 fragments, and iii) ApoA1 or ApoA1 fragments. including implementing. In certain embodiments, detecting the level of at least one HDL-associated protein further comprises adding a labeled or unlabeled ApoA1 protein or protein fragment standard to the sample and detecting the ApoA1 standard. include. In other embodiments, the ApoA1 calibrator comprises or consists of the amino acid sequence set forth in SEQ ID NO:1 or 2.

In further embodiments, the at least one HDL-associated protein further comprises apolipoprotein C1 (ApoC1). In some embodiments, detecting the level of at least one HDL-associated protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments and ii) ApoC1 or ApoC1 fragments. In further embodiments, the at least one HDL-associated protein further comprises apolipoprotein C1 (ApoC2). In some embodiments, detecting the level of at least one HDL-associated protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments and ii) ApoC1 or ApoC2 fragments. In further embodiments, the at least one HDL-associated protein further comprises apolipoprotein C4 (ApoC4). In some embodiments, detecting the level of at least one HDL-associated protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments and ii) ApoC1 or ApoC4 fragments. In certain embodiments, detecting the level of at least one HDL-associated protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments, ii) ApoA1 or ApoA1 fragments, and iii) ApoC1 or ApoC1 fragments. include. In certain embodiments, detecting the level of at least one HDL-associated protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments, ii) ApoA1 or ApoA1 fragments, and iii) ApoC2 or ApoC2 fragments. including.

In some embodiments, the at least one HDL-related protein further comprises apolipoprotein L1 (ApoL1) and phospholipid transfer protein (PLTP). In other embodiments, the detection of the level of at least one HDL-related protein includes i) ApoC3 or ApoC3 fragment, ii) SAA1/2 or SAA1/2 fragment, iii) ApoL1 or ApoL1 fragment, and iv) PLTP or Including performing an assay that detects the PLTP fragment. In other embodiments, the detection of the level of at least one HDL-associated protein includes the addition of labeled or unlabeled ApoL1 protein or protein fragment standards and/or labeled or unlabeled PLTP protein or protein fragment standards to the sample. adding and detecting an ApoL1 standard and/or a PLTP standard. In additional embodiments, the ApoL1 calibrator comprises or consists of the amino acid sequence set forth in SEQ ID NO: 18 or 19 and the PLTP calibrator comprises or consists of SEQ ID NO: 31 or 32.

In other embodiments, the at least one HDL-associated protein further comprises apolipoprotein E (ApoE). In some embodiments, detecting the level of at least one HDL-associated protein includes i) ApoC3 or ApoC3 fragment, ii) SAA1/2 or SAA1/2 fragment, iii) ApoL1 or ApoL1 fragment, iv) PLTP or PLTP fragments, and v) performing assays that detect ApoE or ApoE fragments. In other embodiments, detecting the level of at least one HDL-associated protein further comprises adding a labeled or unlabeled ApoE protein or protein fragment standard to the sample and detecting the ApoE standard. . In a further embodiment, the ApoE calibrator comprises or consists of the amino acid sequence shown in SEQ ID NO:15 or 16.

In certain embodiments, the at least one HDL-related protein further comprises apolipoprotein A1 (ApoA1) and apolipoprotein D (ApoD). In a further embodiment, detection of the level of at least one HDL-related protein includes i) ApoC3 or ApoC3 fragment, ii) SAA1/2 or SAA1/2 fragment, iii) ApoL1 or ApoL1 fragment, iv) PLTP or PLTP fragment. v) ApoE or ApoE fragments; vi) ApoA1 or ApoA1 fragments; and vii) ApoD or ApoD fragments. In some embodiments, detecting the level of at least one HDL-associated protein further comprises adding a labeled or unlabeled ApoD protein or protein fragment standard to the sample and detecting the ApoD standard. include. In certain embodiments, the ApoD calibrator comprises or consists of the amino acid sequence set forth in SEQ ID NO: 13 or 14. In other embodiments, detecting the level of at least one HDL-associated protein further comprises adding a labeled or unlabeled ApoA1 protein or protein fragment standard to the sample and detecting the ApoA1 standard. .

In other embodiments, the at least one HDL-related protein further comprises apolipoprotein M (ApoM) and phospholipid transfer protein (PLTP). In some embodiments, the detection of the level of at least one HDL-related protein includes i) ApoC3 or ApoC3 fragment, ii) SAA1/2 or SAA1/2 fragment, iii) ApoM or ApoM fragment, and iv) PLTP or performing assays that detect PLTP fragments. In a further embodiment, detection of the level of at least one HDL-associated protein includes the addition of a labeled or unlabeled ApoM protein or protein fragment standard and/or a labeled or unlabeled PLTP protein or protein fragment standard to the sample. and detecting ApoM and/or PLTP standards. In additional embodiments, the ApoM standard comprises or consists of the amino acid sequence shown in SEQ ID NO:20 and the PLTP standard comprises or consists of the amino acid sequence shown in SEQ ID NO:31 or 32.

In additional embodiments, the at least one HDL-associated protein further comprises apolipoprotein C1 (ApoC1). In a further embodiment, detecting the level of at least one HDL-associated protein comprises performing assays that detect i) ApoC3 or ApoC3 fragments, ii) SAA1/2 or SAA1/2 fragments, and iii) ApoC1 or ApoC1 fragments. including doing In some embodiments, detecting the level of at least one HDL-associated protein further comprises adding a labeled or unlabeled ApoC1 protein or protein fragment standard to the sample and detecting the ApoC1 standard. include. In certain embodiments, the ApoC1 calibrator comprises or consists of the amino acid sequence set forth in SEQ ID NO:7 or 8.

In additional embodiments, the at least one HDL-associated protein further comprises apolipoprotein D (ApoD). In a further embodiment, detecting the level of at least one HDL-associated protein comprises performing assays that detect i) ApoC3 or ApoC3 fragments, ii) SAA1/2 or SAA1/2 fragments, and iii) ApoD or ApoD fragments. including doing In other embodiments, detecting the level of at least one HDL-associated protein further comprises adding a labeled or unlabeled ApoD protein or protein fragment standard to the sample and detecting the ApoD standard. . In some embodiments, the ApoD calibrator comprises or consists of the amino acid sequence set forth in SEQ ID NO: 13 or 14.

In some embodiments, the at least one HDL-related protein further comprises at least an additional protein selected from the group consisting of CLU, ApoE, CETP, PON1, ApoC1, ApoA2, ApoC2, ApoM, PLTP, and ApoL1 or consist of it. In other embodiments, detection of the level of at least one HDL-associated protein includes i) ApoC3 or ApoC3 fragment, ii) SAA1/2 or SAA1/2 fragment, and iii) at least one additional protein or fragment thereof. Including performing an assay to detect. In certain embodiments, detecting the level of at least one HDL-associated protein includes adding labeled or unlabeled protein or protein fragment standards corresponding to at least one additional protein and corresponding standards. further comprising detecting In further embodiments, the corresponding standards are SEQ ID NOs: 23, 24, 15, 16, 21, 22, 33, 34, 7, 8, 3, 4, 9, 10, 20, 31, 32, 18, and 19 comprises or consists of the amino acid sequence shown in

In certain embodiments, the at least one HDL-associated protein further comprises ApoC1 and ApoC2. In other embodiments, detecting the level of at least one HDL-associated protein comprises performing assays that detect i) ApoC3 or ApoC3 fragments, ii) ApoC1 or ApoC1 fragments, and iii) ApoC2 or ApoC2 fragments. include. In additional embodiments, detection of the level of at least one HDL-associated protein includes adding labeled or unlabeled ApoC1 protein or protein fragment standards to the sample, and/or labeled or unlabeled ApoC2 protein or Further comprising adding protein fragment standards to the sample and detecting ApoC1 and/or ApoC2 standards. In a further embodiment, the ApoC1 calibrator comprises or consists of the amino acid sequence shown in SEQ ID NO:7 or 8 and the ApoC2 calibrator comprises or consists of the amino acid sequence shown in SEQ ID NO:9 or 10. In further embodiments, the at least one HDL-related protein is selected from the group consisting of apolipoprotein C1 (ApoC1) and ApoM, ApoA1, ApoC2, ApoC4, CLU, SAA4, ApoL1, HP, C3, and PLTP. and at least one additional protein. In a further embodiment, detecting the level of at least one HDL-associated protein includes i) ApoC3 or ApoC3 fragment, ii) SAA1/2 or SAA1/2 fragment, iii) ApoC1 or ApoC1 fragment, and iv) at least one Including performing an assay to detect the additional protein or fragment thereof. In a further embodiment, the detection of the level of at least one HDL-related protein comprises the addition of labeled or unlabeled ApoC1 protein or protein fragment standards and/or labeled or unlabeled additional protein or protein fragment standards. , and detecting an ApoC1 standard and/or an additional protein standard. In other embodiments, the ApoC1 standard comprises or consists of the amino acid sequence set forth in SEQ ID NOs: 7 or 8, and the additional protein standards are SEQ ID NOs: 23, 24, 9, 10, 20, 31, 32, 18, comprising or consisting of the amino acid sequence shown in 19, 1, 2, 12, 36, 37, 27, 28, 25, or 26.

In further embodiments, the methods provided herein comprise detecting the level of each HDL-associated protein of at least one of biomarker panel 1-30 in a sample from the subject. Biomarker panels 1-26 are shown in Table 2 and biomarker panels 27-30 are shown in Table 54. In certain embodiments, the subject is human (eg, human male or human female). In certain embodiments, the human has been diagnosed with cardiovascular disease.

In some embodiments, the systems and compositions provided herein comprise: a) a sample from a subject having or suspected of having cardiovascular disease; and b) a first component wherein the first component comprises i) an apolipoprotein C3 (ApoC3) binding agent, and/or ii) an ApoC3 mass spectrometry standard.

In certain embodiments, the systems and compositions c) function as calibrators to normalize the signal in the sample with the approximate level of total HDL particles, ApoA1, or HDL cholesterol present in the sample. It further comprises a second component comprising a composition comprising the detectably labeled HDL protein obtained. In other embodiments, the detectably labeled HDL protein comprises a labeled (eg, isotopically labeled) ApoA1 protein or fragment thereof. In certain embodiments, the sample is selected from the group consisting of serum samples, plasma samples, blood samples, and purified high density lipoprotein (HDL) samples. In certain embodiments, ApoC3 binding agents include anti-ApoC3 antibodies or binding portions thereof, or anti-ApoC3 nucleic acid or protein aptamers or binding portions thereof. In certain embodiments, the ApoC3 mass spectrometry standard comprises isotopically labeled or unlabeled ApoC3 protein or protein fragments. In a further embodiment, the protein fragment comprises or consists of the amino acid sequence shown in SEQ ID NO:11.

In other embodiments, the system or composition further comprises a second component comprising c) i) a serum amyloid A1/2 (SAA1/2) binder, and/or ii) an SAA1/2 mass spectrometry standard. include. In certain embodiments, the SAA1/2 binding agent comprises an anti-SAA1/2 antibody or binding portion thereof, or an anti-SAA1/2 nucleic acid or protein aptamer or binding portion thereof. In some embodiments, SAA1/2 mass spectrometry standards include labeled (eg, isotopically labeled) or unlabeled SAA1/2 proteins or protein fragments. In a further embodiment, the protein fragment comprises or consists of the amino acid sequence shown in SEQ ID NO:35.

In additional embodiments, the system or composition further comprises a third component comprising d) i) an apolipoprotein A1 (ApoA1) binding agent, and/or ii) an ApoA1 mass spectrometry standard. In further embodiments, the ApoA1-binding agent comprises an anti-ApoA1 antibody or binding portion thereof, or an anti-ApoA1 nucleic acid or protein aptamer or binding portion thereof. In a further embodiment, the ApoA1 mass spectrometry standard comprises isotopically labeled or unlabeled ApoA1 protein or protein fragments. In certain embodiments, the protein fragment comprises or consists of the amino acid sequence set forth in SEQ ID NO:1 or 2.

In additional embodiments, the system or composition comprises d) a third component comprising i) an apolipoprotein L1 (ApoL1) binding agent, and/or ii) an ApoL1 mass spectrometry standard; and e) i) a phospholipid transport a fourth component comprising a protein (PLTP) binding agent, and/or ii) a PLTP mass spectrometry standard. In further embodiments, the ApoL1-binding agent comprises an anti-ApoA1 antibody or binding portion thereof, or an anti-ApoL1 nucleic acid or protein aptamer or binding portion thereof, and the PLTP-binding agent comprises an anti-PLTP antibody or binding portion thereof, or an anti- Including PLTP nucleic acid or protein aptamers or binding portions thereof. In certain embodiments, the ApoL1 mass spectrometry standard comprises isotopically or unlabeled ApoL1 protein or protein fragment, and/or the PLTP mass spectrometry standard comprises isotopically or unlabeled PLTP protein or contain protein fragments. In a further embodiment, the ApoL1 protein fragment comprises or consists of the amino acid sequence set forth in SEQ ID NO:18 or 19 and/or the PLTP protein fragment comprises or consists of the amino acid sequence set forth in SEQ ID NO:31 or 32.

In some embodiments, the system or composition further comprises a fifth component comprising i) an apolipoprotein E (ApoE) binding agent, and/or ii) an ApoE mass spectrometry standard. In other embodiments, ApoE binding agents include anti-ApoE antibodies or binding portions thereof, or anti-ApoE nucleic acid or protein aptamers or binding portions thereof. In additional embodiments, the ApoE mass spectrometry standard comprises isotopically labeled or unlabeled ApoE protein or protein fragments. In other embodiments, the ApoE protein fragment comprises or consists of the amino acid sequence set forth in SEQ ID NO:15 or 16.

In certain embodiments, the systems and compositions further comprise a sixth component comprising i) an apolipoprotein A1 (ApoA1) binding agent, and/or ii) an ApoA1 mass spectrometry standard. In other embodiments, the system comprises d) a third component comprising i) an apolipoprotein M (ApoM) binder, and/or ii) an ApoM mass spectrometric standard; and e) i) a phospholipid transfer protein (PLTP ) a binding agent, and/or ii) a fourth component comprising a PLTP mass spectrometry standard. In certain embodiments, ApoM binding agents include anti-ApoM antibodies or binding portions thereof, or anti-ApoM nucleic acid or protein aptamers or binding portions thereof. In certain embodiments, ApoM mass spectrometry standards include isotopically or unlabeled ApoM proteins or protein fragments. In other embodiments, the ApoM protein fragment comprises or consists of the amino acid sequence set forth in SEQ ID NO:20.

In additional embodiments, the systems and compositions further comprise a third component comprising d) i) an apolipoprotein C1 (ApoC1) binding agent, and/or ii) an ApoC1 mass spectrometry standard. In additional embodiments, the ApoC1 binding agent comprises an anti-ApoC1 antibody or binding portion thereof, or an anti-ApoC1 nucleic acid or protein aptamer or binding portion thereof. In other embodiments, the ApoC1 mass spectrometry standards include labeled (eg, isotopically labeled) or unlabeled ApoC1 protein or protein fragments. In certain embodiments, the ApoC1 protein fragment comprises or consists of the amino acid sequence set forth in SEQ ID NO:7 or 8.

In other embodiments, the systems and compositions further comprise a third component comprising d) i) an apolipoprotein D (ApoD) binding agent, and/or ii) an ApoD mass spectrometry standard. In some embodiments, ApoD binding agents include anti-ApoD antibodies or binding portions thereof, or anti-ApoD nucleic acid or protein aptamers or binding portions thereof. In further embodiments, the ApoD mass spectrometry standard comprises labeled (eg, isotopically labeled) or unlabeled ApoD protein or protein fragment. In a further embodiment, the ApoD protein fragment comprises or consists of the amino acid sequence set forth in SEQ ID NO:13 or 14.

In certain embodiments, the systems and compositions further comprise a fourth component comprising d) i) an additional protein binding agent, and/or ii) a mass spectrometry standard of the additional protein, wherein the additional protein is , CLU, ApoE, CETP, PON1, ApoC1, ApoA2, ApoC2, ApoM, PLTP, and ApoL1. In further embodiments, the additional protein binding agent comprises an anti-additional protein antibody or binding portion thereof, or an anti-additional protein nucleic acid or protein aptamer or binding portion thereof. In other embodiments, the additional protein mass spectrometry standards include isotope-labeled or unlabeled additional proteins or protein fragments. In other embodiments, the additional protein fragments are SEQ ID NOs: 23, 24, 15, 16, 21, 22, 33, 34, 7, 8, 3, 4, 9, 10, 20, 31, 32, 18, and 19 comprises or consists of the amino acid sequence shown in

In some embodiments, the systems and compositions comprise c) i) an ApoC1 binding agent, and/or ii) a second component comprising an ApoC1 mass spectrometry standard, and d) i) an ApoC2 binding agent, and/or ii) a third component comprising an ApoC2 mass spectrometry standard; In further embodiments, the ApoC1-binding agent comprises an anti-ApoC1 antibody or binding portion thereof, or an anti-ApoC1 nucleic acid or protein aptamer or binding portion thereof, and the ApoC2-binding agent comprises an anti-ApoC2 antibody or binding portion thereof, or an anti- ApoC2 nucleic acid or protein aptamers or binding portions thereof. In other embodiments, the ApoC1 mass spectrometry standard comprises isotopically or unlabeled ApoC1 protein or protein fragment and the ApoC2 mass spectrometry standard comprises isotopically or unlabeled ApoC2 protein or protein fragment. include.

In certain embodiments, the systems and compositions comprise a third component comprising d) i) an apolipoprotein C1 (ApoC1) binding agent, and/or ii) an ApoC1 mass spectrometry standard, and e) i) an additional protein binding and/or ii) a fourth component comprising mass spectrometric standards of additional proteins, wherein the additional proteins are ApoM, ApoA1, ApoC2, ApoC4, CLU, SAA4, ApoL1, HP, C3, and PLTP. In other embodiments, the ApoC1 binding agent comprises an anti-ApoC1 antibody or binding portion thereof, or an anti-ApoC1 nucleic acid or protein aptamer or binding portion thereof, and the additional protein binding agent comprises an anti-additional protein antibody or binding portion thereof. , or anti-add protein nucleic acids or protein aptamers or binding portions thereof. In some embodiments, the ApoC1 mass spectrometry standard comprises an isotopically or unlabeled ApoC1 protein or protein fragment, and the additional protein mass spectrometry standard comprises an isotopically or unlabeled additional protein or Contains protein fragments. In further embodiments, the ApoC1 protein fragment comprises or consists of the amino acid sequence set forth in SEQ ID NOs: 7 or 8, and additional protein fragments are , 1, 2, 12, 36, 37, 27, 28, 25, or 26. In some embodiments, the subject is human. In certain embodiments, the human has been diagnosed with cardiovascular disease.

In some embodiments, methods are provided herein for determining an estimated risk of cardiovascular disease (e.g., CAD, atherosclerotic disease, etc.) and/or an estimated reverse cholesterol transport capacity, comprising: ) of each HDL-associated protein of at least one of biomarker panels 1-30 (biomarker panels 1-26 are shown in Table 2 and biomarker panels 27-30 are shown in Table 54) in a sample from the subject; b) determining the estimated risk of cardiovascular disease (CVD) in the subject and/or the estimated cholesterol efflux capacity (CEC) of the sample.

In some embodiments, the sample is selected from the group consisting of serum samples, plasma samples, blood samples, and purified high density lipoprotein (HDL) samples. In a further embodiment, the determining comprises generating a cardiovascular disease (CVD) risk score or a cholesterol efflux capacity (CEC) score using a first algorithm, wherein the first algorithm is i a) multiplying each HDL-associated protein level by a predetermined factor to produce a multiplied value; and ii) summing the multiplied values and adding a panel-specific constant value, thereby yielding a CVD risk score. Or generate a CEC score. In other embodiments, the method further comprises c) generating a report indicative of the CVD risk score and/or the CEC score. In certain embodiments, the report further includes levels of at least one HDL-related protein. In other embodiments, the determining further comprises generating a probability of CVD using a second algorithm, wherein the second algorithm calculates the CVD risk score according to the following formula: CVD probability = 1/ Apply to (1+exp(−risk score)) or similar formula. In other embodiments, the method further includes c) generating a report indicating the probability of CVD (eg, CAD). In further embodiments, the report further comprises levels of at least one HDL-related protein.

In certain embodiments, the biomarker panel comprises 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, selected from the group consisting of 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, or 30; In some embodiments, the biomarker panel is biomarker panel number 18, composed of HDL-associated proteins ApoA1, ApoC1, ApoC2, ApoC3, and ApoC4. In other embodiments, the biomarker panel is biomarker panel number 19, composed of HDL-associated proteins ApoA2, ApoC1, ApoC2, ApoC3, ApoD, and SAA1/2. In other embodiments, the biomarker panel is biomarker panel number 5, composed of HDL-associated proteins ApoA1, ApoC2, and ApoC3. In additional embodiments, the biomarker panel is biomarker panel number 4, composed of HDL-associated proteins ApoCl, ApoC3, CLU, PLTP, and SAA4. In another embodiment, the biomarker panel consists of the HDL-associated proteins ApoA1, ApoA2, ApoC1, ApoC2, ApoC3, ApoC4, ApoD, ApoE, ApoL1, ApoM, C3, CLU, HP, SAA1/2, and SAA4 Biomarker panel number 28. In a further embodiment, the biomarker panel is biomarker panel number 30 consisting of HDL-associated proteins ApoA1, ApoA2, ApoC3, ApoC4, ApoD, ApoE, ApoL1, ApoM, C3, HP, PLTP, PON1, and SAA1/2 is. In some embodiments, the CVD is coronary artery disease (CAD). In a further embodiment, the CEC is a global CEC. In a further embodiment, the CEC is the CEC of ABCA1.

In some embodiments, the method further comprises, after step a) but before step b), normalizing the levels of each of the at least one biomarker panel's HDL-related proteins, wherein Normalization is the normalization with the total HDL particles, or ApoA1, or the approximate level of HDL cholesterol in the sample to generate a normalized value for HDL protein. In additional embodiments, the method further comprises determining the estimated level of total HDL particles, or ApoA1, or HDL cholesterol in the sample. In a further embodiment, determining the approximate level of total HDL particles, or ApoA1, or HDL cholesterol in the sample comprises determining the level of an internal standard added to the sample, wherein the internal standard contains labeled HDL protein.

In some embodiments, detecting the level of each of the HDL-related proteins of the at least one biomarker panel includes applying an assay that detects each of the HDL-related proteins of the biomarker panel or fragments thereof to at least a portion of the sample. including implementing In certain embodiments, the assay is a mass spectrometry assay or an immunoassay. In a further embodiment, detecting the level of each HDL-related protein of the at least one biomarker panel includes adding a labeled or unlabeled protein fragment standard to the sample corresponding to at least one of the HDL-related proteins to be detected. and detecting the standard. In other embodiments, the subject is human. In further embodiments, the human has been diagnosed with cardiovascular disease.

In other embodiments, the systems and compositions provided herein comprise a) a sample from a subject having or suspected of having cardiovascular disease, and b) at least one of i) a binding agent to each HDL-associated protein of at least one of biomarker panels 1-30 (biomarker panels 1-26 are shown in Table 2 and biomarker panels 27-30 are shown in Table 54) and/or ii) a mass spectrometry standard protein for each HDL-associated protein of at least one of the biomarker panels 1-30.

In certain embodiments, the system further comprises c) a report indicating raw or normalized levels of HDL-related proteins in the sample for at least one biomarker panel. In a further embodiment, the system further c) reports indicating the subject's CVD risk score and/or cholesterol efflux capacity (CEC) based on the level of HDL-associated proteins in the sample for at least one biomarker panel. include. In other embodiments, the system further includes c) a report indicating the probability risk of CVD in the subject. In certain embodiments, the subject is human.

In some embodiments, provided herein are methods of reporting a CVD risk score or cholesterol efflux capacity score in a subject, comprising: a) Biomarker Panels 1-30 (Biomarker Panels 1-26 are by obtaining a target value for each HDL-associated protein of at least one of the biomarker panels 27-30 shown in Table 2 and shown in Table 54); and b) processing the target value with a processing system. , determining a CVD risk score and/or a cholesterol withdrawal capacity score in a subject, the processing system comprising: i) a computer processor; and ii) a non-transitory computer comprising one or more computer programs and databases. a memory, wherein the one or more computer programs include a biomarker panel model algorithm, the database comprising: i) predetermined coefficients for each HDL-related protein of the at least one biomarker panel; and ii) the at least one biomarker panel. and the one or more computer programs, in conjunction with the computer processor, apply the algorithm of the biomarker panel model, i) for each of the HDL-related proteins of the at least one biomarker panel. multiplying the levels by corresponding predetermined factors to produce multiplied values, ii) summing the multiplied values and adding a panel-specific constant value, thereby the CVD risk score or cholesterol withdrawal capacity (CEC) in the subject configured to generate scores.

In certain embodiments, the method further comprises c) reporting the subject's CVD risk score and/or CEC score as determined by the processing system. In some embodiments, a CVD risk score is used to determine the probability of CVD risk in a subject. In further embodiments, the method further comprises d) performing one or more of the following: i) performing coronary catheterization on the subject on the basis of a high probability of CVD risk; ii) Cardiovascular disease (CVD) medications (e.g., statins, ACE inhibitors, aldosterone inhibitors, angiotensin II receptor blockers, beta-blockers, based on high or intermediate risk of CVD) drugs, calcium channel blockers, cholesterol-lowering drugs, digoxin, diuretics, potassium, magnesium, vasodilators or warfarin); iv) performing at least one additional diagnostic test on the subject based on an intermediate or high probability of CVD risk; v) CVD risk vi) hospitalization and/or ordering the subject on the basis of a high probability of testing a sample from the subject with a CVD risk assay; vii) discharging the subject from the treatment facility on the basis of a low probability of CVD risk; and viii) whether the probability of CVD risk is intermediate. or subject the subject to a stress test on the basis that it is high.

In certain embodiments, the method further comprises d) performing one or more of the following: i) communicating to the user the probability of CVD risk in the subject; ii) determining the probability of CVD risk in the subject; iii) generating a report indicating the probability of CVD risk in the subject; and iv) creating and/or transmitting the report indicating the probability of CVD risk in the subject. In other embodiments, obtaining target values includes receiving target values from a laboratory, from a subject, from an analytical test system, and/or from a handheld or point-of-care test device. In some embodiments, the processing system further includes an analytical test system and/or a handheld or point-of-care testing device.

In certain embodiments, obtaining the target value includes receiving the target value electronically. In other embodiments, obtaining a subject value includes testing a sample from the subject with a detection assay. In further embodiments, the detection assay comprises an immunoassay or mass spectrometry assay. In additional embodiments, the processing system further includes a graphical user interface, and the method further includes entering the target value via the graphical user interface. In certain embodiments, the graphical user interface is part of a device selected from desktop computers, notebook computers, tablet computers, smart phones, and point-of-care analytical devices.

In additional embodiments, the processing system further includes a sample analyzer. In a further embodiment, at least part of the computer memory is located inside the sample analyzer. In further embodiments, the processing system further includes a laboratory interface system (LIM). In other embodiments, at least a portion of the computer memory is part of the LIM.

In some embodiments, the processing system further comprises a processing device selected from the group consisting of desktop computers, notebook computers, tablet computers, smart phones, and point-of-care analysis devices. In additional embodiments, at least a portion of the computer memory is located within the processing unit.

In other embodiments, the one or more computer programs further comprise a CVD probability algorithm, wherein the one or more computer programs, in conjunction with the computer processor, utilize the CVD probability algorithm algorithm, and iii) generate a CVD risk score. It is further configured to apply the following formula: CVD probability = 1/(1 + exp (- CVD risk score). In certain embodiments, the CVD is coronary artery disease (CAD). In other embodiments, The CEC is a global CEC, hi some embodiments, the CEC is an ABCA1 CEC.

In additional embodiments, the target value for each HDL-associated protein is a normalized level (e.g., total HDL particles in the sample, or ApoA1, or an estimated level of HDL cholesterol, or some other protein in the sample, e.g. for approximate levels of HSA). In a further embodiment, the at least one biomarker panel is biomarker panel number 18, composed of HDL-associated proteins ApoA1, ApoC1, ApoC2, ApoC3, and ApoC4. In other embodiments, the at least one biomarker panel is biomarker panel number 19, composed of HDL-associated proteins ApoA2, ApoC1, ApoC2, ApoC3, ApoD, and SAA1/2. In additional embodiments, the at least one biomarker panel is biomarker panel number 5, which is composed of HDL-associated proteins ApoA1, ApoC2, and ApoC3. In other embodiments, the at least one biomarker panel is biomarker panel number 4, composed of HDL-associated proteins ApoCl, ApoC3, CLU, PLTP, and SAA4. In a further embodiment, the at least one biomarker panel consists of the HDL-associated proteins ApoA1, ApoA2, ApoC1, ApoC2, ApoC3, ApoC4, ApoD, ApoE, ApoL1, ApoM, C3, CLU, HP, SAA1/2, and SAA4. Biomarker panel number 28, which is In other embodiments, the at least one biomarker panel is composed of HDL-associated proteins ApoA1, ApoA2, ApoC3, ApoC4, ApoD, ApoE, ApoL1, ApoM, C3, HP, PLTP, PON1, and SAA1/2. Marker panel number 30. In further embodiments, the subject is human. In other embodiments, the human has been diagnosed with cardiovascular disease.

In further embodiments, the processing systems provided herein comprise a) a computer processor and b) a non-transitory computer memory containing one or more computer programs and databases, wherein one or more computer The program includes a biomarker panel model algorithm, and the database contains i) at least of biomarker panels 1-30 (biomarker panels 1-26 are shown in Table 2 and biomarker panels 27-30 are shown in Table 54) ii) a panel-specific predetermined coefficient for each HDL-related protein of at least one of biomarker panels 1-30; iii) for at least one biomarker panel panel-specific constant values, and one or more computer programs, in conjunction with the computer processor, apply the algorithm of the biomarker panel model to: i) subject values for each HDL-related protein to a corresponding panel-specific constant value; multiplying by a predetermined factor to generate a multiplied value; ii) summing the multiplied values and adding a panel-specific constant value to thereby generate a CVD risk score or Cholesterol Evacuation Capacity (CEC) score for the subject; is configured to

In certain embodiments, the one or more computer programs further comprise a CVD probability algorithm, wherein the one or more computer programs, in conjunction with the computer processor, utilize the CVD probability algorithm, and iii) generate a CVD risk score. It is further configured to apply the following formula: CVD probability = 1/(1 + exp (- CVD risk score). In some embodiments, the CVD is coronary artery disease (CAD). In further embodiments, The CEC is a global CEC, hi another embodiment, the CEC is an ABCA1 CEC.

In certain embodiments, the target value for each HDL-associated protein is the normalized level for total HDL particles, or ApoA1, or HDL cholesterol in the sample. In other embodiments, the system further comprises c) an HDL-related protein assay test system and/or a handheld or point-of-care HDL-related protein point-of-care test device. In certain embodiments, the HDL-associated protein assay test system includes a mass spectrometer or an optical detector. In other embodiments, the system further comprises c) a graphical user interface for entering target values for each of the HDL-related proteins into the computer memory. In other embodiments, the graphical user interface is part of a device selected from desktop computers, notebook computers, tablet computers, smart phones, and point-of-care analytical devices.

In some embodiments, the system further includes a sample analyzer, and at least a portion of the computer memory is located within the sample analyzer. In certain embodiments, the system further includes at least part of a laboratory interface system (LIM). In other embodiments, at least a portion of the computer memory is part of the LIM. In other embodiments, the system further comprises c) a processing device selected from the group consisting of desktop computers, notebook computers, tablet computers, smart phones, and point-of-care analyzers. In other embodiments, at least a portion of the computer memory is located within the processing unit. In additional embodiments, the system further includes a display component configured to display the patient's CVD risk score or CEC score. In other embodiments, the display component is selected from computer monitors, tablet computer screens, smartphone screens, and point-of-care analyzer screens.

In some embodiments, the at least one biomarker panel is biomarker panel number 18, composed of HDL-associated proteins ApoA1, ApoC1, ApoC2, ApoC3, and ApoC4. In a further embodiment, the at least one biomarker panel is biomarker panel number 19, composed of HDL-associated proteins ApoA2, ApoC1, ApoC2, ApoC3, ApoD, and SAA1/2. In other embodiments, the at least one biomarker panel is biomarker panel number 5, which is composed of HDL-associated proteins ApoA1, ApoC2, and ApoC3. In a further embodiment, the at least one biomarker panel is biomarker panel number 4, composed of HDL-associated proteins ApoCl, ApoC3, CLU, PLTP, and SAA4. In some embodiments, the at least one biomarker panel is HDL-associated proteins ApoA1, ApoA2, ApoC1, ApoC2, ApoC3, ApoC4, ApoD, ApoE, ApoL1, ApoM, C3, CLU, HP, SAA1/2, and SAA4 Biomarker panel number 28, consisting of In certain embodiments, the at least one biomarker panel is composed of HDL-associated proteins ApoA1, ApoA2, ApoC3, ApoC4, ApoD, ApoE, ApoL1, ApoM, C3, HP, PLTP, PON1, and SAA1/2. Marker panel number 30. In other embodiments, the subject is human. In additional embodiments, the human has been diagnosed with cardiovascular disease.

In some embodiments, a non-transitory computer memory component provided herein comprises one or more computer programs configured to access a database, the one or more computer programs comprising a biomarker panel model algorithm, wherein the database includes i) at least one of biomarker panels 1-30 (biomarker panels 1-26 are shown in Table 2 and biomarker panels 27-30 are shown in Table 54); , a target value for each HDL-associated protein, ii) a predetermined coefficient for each HDL-associated protein of at least one of the biomarker panels 1-30, and iii) a panel-specific constant value for at least one biomarker panel. and wherein one or more computer programs, in conjunction with the computer processor, apply the algorithm of the biomarker panel model to: i) multiply the target value for each HDL-related protein by a corresponding predetermined factor to obtain the multiplied value ii) sum the multiplied values and add a panel-specific constant value, thereby generating a CVD risk score or cholesterol efflux capacity (CEC) score for the subject.

In further embodiments, assays for HDL-associated proteins are from assays including mass spectrometry (MS), chromatography, LC-MS, plasmon resonance, and the use of polyvinyl sulfonic acid (PVS) and polyethylene glycol methyl ether (PEGME). performed using a technique selected from the group consisting of: In certain embodiments, detection is performed by injecting a purified HDL sample into an instrument that performs both chromatography and mass spectrometry. In some embodiments, the device is a liquid chromatography-mass spectrometry (LC/MS) device.

In certain embodiments, the HDL values are normalized values. In certain embodiments, normalization is to total HDL cholesterol, or ApoA1 protein, or HDL particles. Determination of total HDL can be performed by measuring HDL cholesterol. Generally, in practice, a "homogeneous" assay using select reagents added in a specific order is used to "deplete" a serum sample of LDL-cholesterol particles containing lipoprotein ApoB. HDL cholesterol is then determined chemically using a conventional enzyme-linked assay. Measurement of total HDL can also be performed using physical methods to isolate HDL particles, usually ultracentrifugation (e.g. Warnick et al., Clinical Chemistry September 2001
vol. 47 no. 9 1579-1596, incorporated herein by reference). In some embodiments, the measured amount of HDL-associated protein may be compared to the total amount of native ApoA1 in the original sample to determine a normalized ratio. ApoA1 is the major lipoprotein component of each HDL particle. Measurement of HDL cholesterol, rather than ApoA1, has been the mainstay of cardiovascular risk assessment, but this view is changing as measurement of ApoA1 is useful in identifying asymptomatic atherosclerosis (Florvall et al. ., Journal of Gerontology: BIOLOGICAL SCIENCES 2006, Vol. 61A, No. 12, 1262-1266, incorporated herein by reference). Total ApoA1 is typically measured using a widely available immunoassay platform assay.

In some embodiments, methods provided herein include at least three (or at least two) HDL-related proteins (e.g., those listed in Tables 2 and 54) in a sample from a subject. ), wherein the subject is suffering from or suspected of suffering from cardiovascular disease. In certain embodiments, the sample is a purified high-density lipoprotein sample.

In certain embodiments, provided herein are methods of determining an estimated risk of cardiovascular disease and/or an estimated reverse cholesterol transport capacity, comprising: a) at least three species in a sample from a subject ( or at least two) HDL-related proteins (e.g., selected from those listed in Table 2 and Table 54); b) an estimated risk of cardiovascular disease (CVD) in the subject; and/or determining the approximate cholesterol efflux capacity (CEC) of the sample.

In some embodiments, the systems provided herein comprise a) a sample from a subject having or suspected of having cardiovascular disease, and b) at least one of and/or ii) at least three (or at least two) mass spectrometry standard proteins for each of the HDL-related proteins.

In other embodiments, provided herein are methods of reporting a CVD risk score or cholesterol efflux capacity score in a subject, comprising: a) at least three (or at least two) HDL-related proteins (e.g. , selected from those listed in Table 2 and Table 54); determining an extractability score, the processing system comprising: i) a computer processor; ii) a non-transitory computer memory comprising one or more computer programs and a database; The program includes an algorithm of biomarkers, wherein the database includes: i) predetermined coefficients for each of at least three (or at least two) HDL-related proteins; and one or more computer programs, in conjunction with a computer processor, apply a biomarker algorithm to i) each of at least three (or at least two) HDL-associated proteins; ii) sum the multiplied values and add a panel-specific constant value, thereby yielding a CVD risk score or cholesterol withdrawal capacity (CEC ) is configured to generate a score.

In some embodiments, a processing system provided herein comprises a) a computer processor and b) non-transitory computer memory containing one or more computer programs and databases, wherein one or more comprises a biomarker algorithm, wherein the database i) targets for each of at least three (or at least two) HDL-associated proteins (e.g., selected from those listed in Table 2 and Table 54) ii) a panel-specific predetermined coefficient for each of the at least three (or at least two) HDL-related proteins; and iii) a panel-specific constant value for a combination of the at least three HDL-related proteins; The one or more computer programs, in conjunction with the computer processor, apply the biomarker algorithm to: i) target values for each of the at least three (or at least two) HDL-associated proteins; configured to multiply the coefficients to generate a multiplied value, ii) sum the multiplied values and add a panel-specific constant value, thereby generating a CVD risk score or cholesterol efflux capacity (CEC) score for the subject It is

In certain embodiments, a non-transitory computer memory component provided herein comprises one or more computer programs configured to access a database, the one or more computer programs comprising a marker algorithm, wherein the database contains: i) a target value for each of at least three (or at least two) HDL-related proteins (e.g., selected from those listed in Tables 2 and 54); and ii) iii) a panel-specific constant value for a combination of at least three (or at least two) HDL-related proteins, wherein one or more computer programs are associated with the computer processor; and apply the biomarker's algorithm to i) multiply the target value for each HDL-associated protein by a corresponding predetermined factor to generate a multiplied value, and ii) sum the multiplied values to yield a panel-specific constant value. , thereby generating a CVD risk score or cholesterol efflux capacity (CEC) score for the subject.

In certain embodiments, the biomarker panels described herein are for cardiovascular disease, non-alcoholic steatohepatitis (NASH), lupus erythematosus, irritable bowel syndrome (IBS), chronic kidney disease (CKD), It is used to diagnose the risk of HDL-related diseases selected from rheumatoid arthritis (RA) and Alzheimer's disease.

The present invention can also be configured as follows.
[1] A method comprising detecting the level of at least one HDL-related protein in a sample from a subject, wherein the at least one HDL-related protein comprises ApoC3, and the subject suffers from cardiovascular disease having or suspected of having the disease.
[2] The method of [1], wherein the sample is a purified high-density lipoprotein sample.
[3] The method of [1], wherein the sample is selected from the group consisting of serum samples, plasma samples, and blood samples.
[4] normalizing said level of said at least one HDL-associated protein with said estimated level of total HDL particles, or ApoA1, or HDL cholesterol to produce a normalized value of at least one HDL protein; The method of [1], further comprising:
[5] said determination of the estimated level of total HDL particles, or ApoA1, or HDL cholesterol in said sample includes determining the level of an internal standard added to said sample, wherein said internal standard is labeled; The method of [4], comprising HDL protein.
[6] The method of [1], wherein said detecting the level of at least one HDL protein comprises adding a reagent to said sample, said reagent digesting HDL protein in said sample.
[7] The method of [1], wherein said detecting the level of at least one HDL-associated protein comprises performing an assay that detects ApoC3 or ApoC3 fragments on at least a portion of said sample.
[8] The method of [7], wherein the assay is a mass spectrometry assay or an immunoassay.
[9] The method of [1], wherein the at least one HDL-related protein further comprises apolipoprotein A1 (ApoA1).
[10] The method of [9], wherein said detecting the level of at least one HDL protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments, and ii) ApoA1 or ApoA1 fragments.
[11] The method of [1], wherein the at least one HDL-related protein further comprises apolipoprotein C1 (ApoC1).
[12] The method of [11], wherein said detecting the level of at least one HDL-associated protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments, and ii) ApoC1 or ApoC1 fragments.
[13] The method of [1], wherein the at least one HDL-related protein further comprises apolipoprotein C2 (ApoC2).
[14] The method of [13], wherein said detecting the level of at least one HDL-associated protein comprises performing an assay that detects i) ApoC3 or ApoC3 fragments, and ii) ApoC2 or ApoC2 fragments.
[15] a) a sample from a subject suffering from or suspected of suffering from cardiovascular disease;
b) a first component, said first component comprising i) an apolipoprotein C3 (ApoC3) binding agent, and/or ii) an ApoC3 mass spectrometry standard.
[16] c) a detectably labeled HDL protein that can serve as a calibrator for normalizing the signal in said sample with the estimated level of total HDL particles, ApoA1, or HDL cholesterol present in said sample; The system or composition of [15], further comprising a second component comprising the containing composition.
[17] The system or composition of [15], wherein the sample is selected from the group consisting of serum samples, plasma samples, blood samples, and purified high density lipoprotein (HDL) samples.
[18] c) The system or composition of [15], further comprising a second component comprising i) an apolipoprotein A1 (ApoA1) binding agent, and/or ii) an ApoA1 mass spectrometry standard.
[19] c) The system or composition of [15], further comprising a second component comprising i) an apolipoprotein C1 (ApoC1) binding agent, and/or ii) an ApoC1 mass spectrometry standard.
[20] c) The system or composition of [15], further comprising a second component comprising i) an apolipoprotein C2 (ApoC2) binding agent, and/or ii) an ApoC2 mass spectrometry standard.
[21] A method for determining an estimated risk of cardiovascular disease and/or an estimated reverse cholesterol transport capacity, comprising:
a) in a subject-derived sample, biomarker panel 17, 18, and/or 19, wherein biomarker panel 17 comprises proteins ApoC1, ApoC2, and ApoC3;
said biomarker panel 18 comprises proteins ApoA1, ApoC1, ApoC2, ApoC3, and ApoC4;
The biomarker panel includes proteins ApoA2, ApoC1, ApoC2, ApoC3, ApoD, and SAA1/2)
detecting the level of each of the HDL-associated proteins of
b) determining the estimated risk of cardiovascular disease (CVD) in the subject and/or the estimated cholesterol efflux capacity (CEC) of the sample.
[22] The method of [21], wherein the sample is selected from the group consisting of serum samples, plasma samples, blood samples, and purified high density lipoprotein (HDL) samples.
[23] said determining comprises using a first algorithm to generate a cardiovascular disease (CVD) risk score or a cholesterol efflux capacity (CEC) score, said first algorithm comprising: i) each HDL-related multiplying the protein level by a predetermined factor to produce a multiplied value, and ii) summing said multiplied values and adding a panel-specific constant value, thereby obtaining said CVD risk score or said A method according to [21] for generating a CEC score.
[24] c) The method of [21], further comprising generating a report indicating the CVD risk score and/or the CEC score.

1 shows an exemplary workflow corresponding to the steps described in Example 1 below for Panel 1. FIG. Isolation of ApoA-I related serum fractions—ApoA-I related proteins were isolated by incubating serum with His ₆ ApoA-I and targeting the His-tag using immobilized metal affinity chromatography (IMAC). Schematic diagram of release. Overlap of ApoA-I-associated proteome and HDL proteome—A Venn diagram showing the overlap of proteins identified by ApoA-I affinity purification and HDL prepared by gradient ultracentrifugation within the HDL Proteome Watch list. Implications for Cell-Based Cholesterol Extraction Assays—Based on measured cholesterol efflux capacity (CEC) from cAMP-stimulated J774 macrophages and (A) targeted analysis of ApoA-I-associated serum proteins using a panel of 6 proteins. predicted CEC and correlation with serum concentrations of (B) ApoA-I, (C) HDL, and (D) hsCRP. Inverse Correlation between Predicted CEC and CAD—Distribution of CEC predicted by targeted HDL proteome analysis. Comparison of total control and CAD patient populations (with or without MACE). Correlation between endogenous ApoA-I observed in targeted MRM measurements of ApoA-I-related proteins and serum ApoA-I levels measured by immunonephelometry in a referral clinical laboratory. CEC predicted from targeted HDL proteomic analysis of 30 samples (n=15 control, n=15 CAD) and observed cellular cholesterol efflux by cell-based assay of the same sample set in cAMP-stimulated J774 macrophages. Compare. Figure 3 shows results from Example 4 comparing predicted cholesterol withdrawal values between cohorts and demonstrating overall lower cholesterol withdrawal in the CAD cohort compared to the control cohort. FIG. 10 shows that application of the outcome optimization model of Example 4 stratifies patients on a biomarker score scale that correlates with risk of being diagnosed with CAD.

DEFINITIONS As used herein, “high-density lipoprotein” or “HDL” is a non-covalent complex of amphipathic proteins that circulates in the blood and transports lipids such as cholesterol and triglycerides to the aqueous blood. Make it possible to transport it in the stream. HDL is composed approximately 50% by mass of an embedded phospholipid monolayer (approximately 25%) of free cholesterol (approximately 4%) and a core of triglycerides (approximately 3%) and cholesterol esters (approximately 12%). It is composed of amphiphilic proteins that stabilize lipid emulsions. Subclasses of HDL include HDL2 and HDL3. HDL2 particles are larger and have a higher lipid content, whereas HDL3 particles are smaller and have a lower lipid content. Other subclasses include HDL2b, HDL2a, HDL3a, HDL3b, and HDL3c, in order from largest particle to smallest particle.

As used herein, "lipoprotein" refers to a class of proteins that bind or are capable of binding one or more lipid molecules. In some cases, the lipoprotein can be a "lipid-deficient lipoprotein" with four or fewer phospholipid molecules attached. As used herein, lipoproteins include proteins that are not lipid bound but are exchangeable in HDL particles (eg, apolipoproteins).

As used herein, the phrase "reverse cholesterol transport" refers to the multi-step process that results in the net transfer of cholesterol from peripheral tissues back to the liver through the plasma.

As used herein, the phrase "cholesterol efflux capacity (CEC)" of a sample refers to the ability of HDL in a sample to facilitate reverse cholesterol transport by accepting cholesterol from lipid-laden macrophages. Methods for measuring CEC include, but are not limited to, de la Llera-Moya et al. , Arterioscler Thromb Vasc Biol. 2010 Apr;30(4):796-801; Sankaranarayanan et al. , J Lipid Res. 2011 Dec;52(12):2332-40; and Khera et al. , NEJM, 364; 2, 2011, all of which are incorporated herein by reference in their entirety, including assays disclosed therein. ABCA1 CEC specifically refers to reverse cholesterol transport specifically induced by the ABCA1 transporter. For example, this can be measured by subtracting the difference in CEC between reagent-treated (eg, cAMP-treated) and reagent-untreated macrophages. In this case, macrophages are denatured such that ABCA1 expression immediately increases in the presence of the treatment reagent.

As used herein, "blood sample" refers to a whole blood sample, or plasma or serum fractions derived therefrom. In certain embodiments, a blood sample refers to a human blood sample, such as a whole blood sample, or plasma or serum fractions derived therefrom. In some embodiments, a blood sample refers to a non-human mammalian (“animal”) blood sample, such as a whole blood sample, or a plasma or serum fraction derived therefrom.

As used herein, the term "whole blood" refers to a blood sample that has not been fractionated and contains both cellular and fluid components.

As used herein, "plasma" refers to the fluid, non-cellular component of whole blood. Depending on the separation method used, the plasma may contain no cellular components, or may contain varying amounts of platelets and/or small amounts of other cellular components. The term "plasma" is distinguished from "serum" described below, as plasma contains various clotting factors such as fibrinogen.

As used herein, the term "serum" refers to whole mammalian serum, eg, whole human serum, whole serum from test animals, whole pet serum, whole livestock serum, and the like. Additionally, as used herein, "serum" refers to plasma from which clotting factors (eg, fibrinogen) have been removed.

As used herein, the phrase "purified high-density lipoprotein sample" refers to a blood sample (e.g., serum, or plasma, or whole blood sample) (e.g., by ultracentrifugation as described below). at least 90% (e.g., at least 90% ... 94% ... 98% ... 99% ... or at least 99.9%) of the total protein in the purified sample (obtained by purification or ApoA1 exchange method) is HDL liposome; Refers to purified samples that are proteins. In some embodiments, less than 10% of the total proteins in the purified sample are non-HDL lipoproteins (eg, less than 10%...5%...1%...0.2%). In certain embodiments, the non-HDL lipoproteins are primarily or entirely serum albumin.

As used herein, the term "cardiovascular disease" (CVD) or "cardiovascular disorder" refers to a number of diseases that affect the body's heart, heart valves, and vascular system (e.g., veins and arteries). A term used to classify conditions, the diseases and conditions included therein include arteriosclerosis, atherosclerosis, myocardial infarction, acute coronary syndrome, angina pectoris, congestive heart failure, aortic aneurysm, aortic dissection, iliac or femoral aneurysm, pulmonary embolism, primary hypertension, atrial fibrillation, stroke, transient ischemic attack, systolic dysfunction, diastolic dysfunction, myocarditis, atrial tachycardia, Ventricular fibrillation, endocarditis, arteritis, vasculitis, atherosclerotic plaque, vulnerable plaque, acute coronary syndrome, acute ischemic attack, sudden cardiac death, peripheral vascular disease, coronary artery disease (CAD), peripheral arterial disease ( PAD), and cerebrovascular disease.

As used herein, the phrase "suspected to have CVD" refers to at least one sign or symptom associated with CVD, e.g., extreme fatigue, constant dizziness or lightheadedness, rapid heart rate (e.g., beats >100 beats per minute at rest), new irregular heartbeat, chest pain or discomfort during activity that disappears with rest, dyspnea during normal activity and at rest, worsening respiratory infection or cough, Refers to patients with anxiety or confusion, altered sleep patterns, and anorexia or nausea.

As used herein, the terms “arteriosclerotic cardiovascular disease” or “arteriosclerotic cardiovascular disorder” refer to the portion or specific type of cardiovascular disease that includes atherosclerosis as a component. including but not limited to CAD, PAD, cerebrovascular disease. Atherosclerosis is a chronic inflammatory reaction that occurs in arterial blood vessel walls. This is accompanied by the formation of atherosclerotic plaques, which can lead to narrowing of the artery (“stenosis”) and ultimately to partial or complete closure of the arterial opening and/or plaque rupture. Thus, arteriosclerotic diseases or disorders include the consequences of atherosclerotic plaque formation and rupture, including stenosis or narrowing of arteries, heart failure, formation of aneurysms, including aortic aneurysms, aortic dissection, and myocardial Including, but not limited to, ischemic events such as infarction and stroke. In certain embodiments, the methods, compositions, and systems disclosed herein are used to at least partially diagnose atherosclerotic CVD.

The terms "individual," "host," "subject," and "patient" are used interchangeably herein and generally include primates, including apes and humans, equines (e.g., horses), canines ( dogs), felines, various farm animals (e.g., ungulates such as pigs, pigs, goats, sheep, etc.), as well as domestic pets and zoo-controlled animals. It refers to non-limiting mammals. In some embodiments, the subject is specifically a human subject.

DETAILED DESCRIPTION OF THE INVENTION Herein, one or more HDL-related proteins (e.g., ApoC3; ApoC3; and ApoA1; ApoC3 and SAA1/2; or proteins of biomarker panels 1-30). In certain embodiments, such methods, systems, and compositions are used to determine the approximate risk of CVD (or other disease) in a subject and/or the approximate cholesterol efflux capacity (CEC) of a sample. do. In certain embodiments, the systems and compositions comprise a sample from a subject having or suspected of having CVD and an HDL-related binding agent or mass spectrometry standard.

Provided herein are methods, compositions, and methods for determining predicted cardiovascular disease and/or cholesterol withdrawal based on measuring protein levels in one or more of the panels shown in Tables 2 and 54 below. and provide kits. From studies conducted during the development of embodiments of the present disclosure, CVD either at the level of a single protein (e.g. panel 10, ApoC3) or multiple protein combinations (e.g. panels 1-9 and 11-30) and the ability to extract cholesterol. Described below are exemplary, non-limiting methods and systems for use in using biomarker panels in samples from subjects suffering from or suspected of suffering from cardiovascular disease. .

In some embodiments, blood is first drawn from a subject with symptoms of cardiovascular disease (eg, chest pain) at a hospital, clinic, or other medical facility. This blood sample is processed to produce a serum sample. The serum sample is then treated with a method that purifies the HDL in the sample, such as ultracentrifugation or purification of tagged HDL-binding particles. Additionally, the purified serum sample is treated with an enzyme (eg, Lys-C, etc.) that digests HDL-associated proteins present in the purified sample. This digested sample is then subjected to one or more detection assays (eg, mass spectroscopy, immunological assays, aptamer binding assays) to determine HDL-associated protein levels present in one or more of biomarker panels 1-30. to detect

In some embodiments, when using mass spectrometry, the detection protocol can be as follows. A portion of the digested HDL purified sample is diluted and loaded onto the LC column. Peptides are then detected using a triple quadrupole mass spectrometer (eg, operating in dynamic MRM mode). Consists of a given m/z value for the mass of the intact peptide (filtered by instrument Q1), the collision energy the intact peptide undergoes by instrument q2, and the fragment (m/z) value for the peptide fragment (filtered by Q3). Peptides are targeted at predetermined 'transitions'. Transitions are selected to be specific to the target peptide within the sample. Two transitions (a "quantifier" used for quantification in the second half and a "qualifier" used for quality control) are monitored per peptide, targeting a maximum of two peptides per protein. A detailed list of peptide targets and their transitions is provided in Table 1 of Example 1 below.

Peptide signal intensities are obtained by integrating the chromatographic peaks of the 'quantifier' transitions for each peptide. Protein intensities are determined by summing the peak areas of the quantifiers for each of the protein's target peptides and normalizing with the intensity of one or more HDL-associated proteins.

In addition, the levels of HDL-related proteins of the selected biomarker panel are processed with a multivariate algorithm to determine cholesterol withdrawal or CVD risk scores. The algorithm uses a constant (i) specific to each panel (and specific to CVD risk score or overall or ABCA1 withdrawal) and a protein- and panel-specific coefficient (c )including. The algorithm is as follows:
CVD risk score/predicted _overall cholesterol withdrawal ₌ c1 ^* p1+c2 _* p2... ^{cn *} _{pn + i}
Here, withdrawal or CVD risk is determined by adding a constant (i) to the additive sum of the factor (c) multiplied by the protein level (p) for a given protein (e.g., normalized peak area). be done. Example 1 below shows the coefficients and constants for each protein in each panel for both protein and molar protein values determined by mass spectrometry. Example 1 further provides 95% confidence intervals for each of these numbers. For example, any value within this range can be used for the value of (c) or (p) when calculating a CVD risk score and/or cholesterol withdrawal value.

The CVD risk score determined using the algorithm above and the values in Example 1 below can be converted to a probability (risk percentage of CVD) using the following formula:
Probability=1/(1+exp(−risk score)).
In addition, this probability value can be multiplied by 100 to obtain a CVD risk rate. A CVD risk score or probability or rate may be used by health care professionals to help determine treatment of patients with suspected heart disease or to exclude patients who may require treatment or monitoring. can be done. For example, in general, a patient with a risk rate of less than 10% for CAD is typically considered low risk (e.g., can be considered a patient who does not require treatment or monitoring for heart disease) and a risk rate for CAD of 10%. Patients at ~20% risk are usually considered intermediate risk (e.g., patients can be monitored for additional signs of heart disease), and patients with >20% risk of CAD are usually considered high risk. considered a risk (eg, can be treated with therapeutic or surgical intervention).

In certain embodiments, one or more of the proteins detected as part of panels 1-30 are detected in the purified high-density lipoprotein sample. Purified HDL samples can be obtained by any suitable method, including ultracentrifugation or HDL-binding peptide methods.

In the case of ultracentrifugation, for example, ultracentrifugation (105,000 xg) of serum with a density of 1.006 kg/L at 18°C for 18 hours is used to prevent sedimentation of low-density lipoproteins by manganese and heparin. Low density lipoproteins can be removed. Manganese chloride and heparin are added to the lower fraction in amounts sufficient to obtain a final concentration of 0.0456 molar Mn ²⁺ and 183 UPS kU/L of heparin to precipitate non-HDL-C lipoproteins. After centrifugation at 4° C. and 1500×g for 30 minutes, the resulting supernatant is further adjusted with potassium bromide to a density of 1.21 kg/L or higher. The density-adjusted solution is then centrifuged again (approximately 105,000 xg) for 24 hours. After centrifugation, careful removal of the top layer can yield an enriched fraction of HDL.

In certain embodiments, HDL is purified using an affinity-tagged HDL-binding peptide (eg, ApoA1). Such purification methods are described in U.S. Application Serial No. 14/713,046, filed May 15, 2015, which is fully incorporated herein by reference. incorporated herein by. Briefly, an affinity-tagged HDL-binding peptide (eg, ApoA1) is added to the sample. HDL particles incorporate a portion of the bound particles. Such HDL particles are then purified utilizing the affinity tag, thereby obtaining a purified HDL sample.

The present invention is not limited to methods used to detect HDL-related peptides (eg, ApoC3, SAA1/2, ApoE, ApoL1, etc.) from subject samples.

In certain embodiments, HDL-associated proteins are detected using a detection method selected from: immunoassays, surface plasmon resonance, in vitro assays, activity assays, co-immunoprecipitation assays, mass spectrometry, fluorescence energy transfer. (FRET), bioluminescence energy transfer (BRET), interferometry, biolayer interferometry (BLI), dual polarization interferometry (“DPI”), ellipsometry, and quartz crystal microbalance.

Example 1
This example demonstrates the ability to predict global cholesterol withdrawal and cardiovascular disease risk by 26 panels of single and multiple HDL proteins using different algorithms for overall cholesterol withdrawal and coronary artery disease risk. Explain testing. HDL protein was measured by mass spectrometry.

Purification of HDL The following protocol was used for purification of HDL. Mix 25 μL of a 0.5 mg/mL solution of ¹⁵ N-labeled His-tagged apolipoprotein AI with 12 μL of human serum in 1× Phosphate Buffered Saline (PBS), pH 7.4. His-tagged ApoA-I is incorporated into HDL particles in the serum sample during incubation of the mixture of serum and ApoA-I at 37° C. for 15 minutes. HDL is then purified from the mixture by immobilized metal affinity chromatography in a pipette tip format suitable for high throughput automated liquid handling platforms. Briefly, 164 μL of binding buffer (5 mM imidazole, 50 mM sodium phosphate, 300 mM sodium chloride, pH 8.0) is added to dilute the serum and ApoA-I mixture to 200 μL. Equilibrate the affinity column tip with 400 μL binding buffer. The affinity column is then used to aspirate 180 μL of the serum/ApoA-I mixture six times at a flow rate of 200 μL/min and dispense onto the column bed. The column bed was then washed once by aspirating and dispensing 200 μL of binding buffer followed by 2 with 200 μL of strong wash buffer (10 mM imidazole, 50 mM sodium phosphate, 300 mM sodium chloride, pH 8.0). Wash for the second time. Bound HDL particles are then eluted by aspirating and dispensing 90 μL of elution buffer (300 mM imidazole, 50 mM Tris-HCl, 25% methanol, pH 9.0).

Digestion of HDL Next, 5 μL of 100 mM dithiothreitol and 10 μL of 50 ng/μL endoproteinase Lys-C are added to the eluted HDL and incubated at 37° C. for 4 hours. Lys-C cleaves HDL-associated proteins C-terminal to the lysine amino acid residues in the sequence to produce predictable target peptides by LC-MS analysis and quantification.

LC-MS Analysis of HDL Proteome Target Peptides 25 μL of Lys-C digested HDL in elution buffer containing 25% methanol was injected in-line using a mixing tee and diluted 5-fold, then directly onto the LC column. to load. Dilution and filling were performed using a sample filling pump that pumped 0.1% formic acid aqueous solution to the mixing tee at 150 μL/min, and 99% mobile phase A (0.1% formic acid aqueous solution), 1% mobile phase B (0.1% % formic acid in acetonitrile) to the tee at 600 μL/min using a binary pump. After loading, the binary pump is switched in-line with the column and the peptides are eluted with a linear gradient of mobile phase B from 1% to 65% at 500 μL/min for 5 minutes. Peptides are detected using an Agilent 6490 triple quadrupole mass spectrometer operating in dynamic MRM mode. Dynamic MRM allows targeted detection and quantification of peptide targets within a scheduled retention window. Consists of a given m/z value for the mass of the intact peptide (filtered by instrument Q1), the collision energy the intact peptide undergoes by instrument q2, and the fragment (m/z) value for the peptide fragment (filtered by Q3). Peptides are targeted at predetermined 'transitions'. Transitions are selected to be specific to the target peptide within the sample. Two transitions (a "quantifier" used for quantification in the second half and a "qualifier" used for quality control) are monitored per peptide, targeting a maximum of two peptides per protein. A detailed list of peptide targets and their transitions is shown in Table 1. It should be noted that moieties other than the listed peptide sequences can be used when performing mass spectrometric measurements.

Panel of 26 HDL Proteins Shown in Table 2 is a panel of 26 proteins showing protein combinations that gave good results in measuring HDL global cholesterol withdrawal.

Peptide and Protein Quantification and Normalization Peptide signal intensities are obtained by integrating the chromatographic peaks of the 'quantifier' transitions for each peptide. Protein intensities are determined by summing the peak areas of the quantifiers for each of the protein's target peptides and normalizing with the intensity of one or more HDL-associated proteins.

Algorithms for Extraction Capacity and CAD Risk Based on intensity normalized from the measured proteins, multiples of extraction capacity (exploration cohort and fresh vs. frozen) and CAD risk (aggregate, with or without event) were calculated. A randomized algorithm was developed and utilized to obtain values related to cholesterol abstraction capacity or CAD risk of original serum samples. Model 1 is global cholesterol withdrawal of exploratory samples, model 2 is global cholesterol withdrawal of fresh and frozen samples, model 3 is total coronary artery disease (CAD) risk, model 4 is CAD risk with events and Model 5 is CAD risk without events. The general formulas for these models are given below:
Models 1 and 2—Predicted overall cholesterol withdrawal=c ₁ ^* p ₁ +c ₂ ^* p ₂ . . .c _n p _n +i
Model 3-5 - CAD (with/without event) risk score = c ₁ p ₁ + c ₂ p ₂ … c _n p _n + i
Here, withdrawal is determined by adding a constant (i) to the additive sum of the coefficient (c) multiplied by the normalized peak area (p) of several proteins.

Statistical Analysis Statistical analysis of the data for each panel was performed using R (“www.” followed by “r-project.org”) and Bioconductor (“www.” followed by “bioconductor.org”). Data were normalized to the total intensity of 15N ApoA1 protein. For proteins with two peptides measured by MRM, protein levels were calculated based on the peptide showing the highest intensity in the majority of samples analyzed. Robust linear regression was used to calculate the correlation between each protein and overall cholesterol withdrawal. The Least absolute shrinkage and selection operator (LASSO) was applied to all proteins and features with non-zero coefficients were selected. Stepwise robust linear regression was applied to selected features measured on 70 exploratory samples. For each panel, the model with the lowest Akaike Information Criterion (AIC) was selected.

Determination of Absolute Molar Amounts of Protein To obtain molar values, a 5-point external calibration curve covering the range of measured signals for each peptide in Table 1 was generated and monitored by LC-MS using synthetic peptide standards. . A standard curve was linearly fitted and weighted by 1/X (X is peptide concentration). The calculated molar values were then used to calculate the molar mass factor.

panel 1
To predict panel 1's overall cholesterol withdrawal, models 1 and 2 were generated. This model is shown below.
Models 1 and 2—Predicted global cholesterol withdrawal=(i)+(c1) ^* ApoC3+(c2) ^* ApoE+(c3) ^* ApoL1+(c4) ^* PLTP+(c5) ^* SAA1/2
The values for Models 1-5 are shown in Table 3A below. As an example, Model 1 has the following values:
Model 1 - predicted global cholesterol withdrawal = 9.25 + (165.75) ApoC3 + (-834.8) ApoE + (-144.26) ApoL1 + (-11548.33) PLTP + (-160.58) SAA1/ 2.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.78. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown above (Table 2) and yielded a correlation of 0.52 between laboratory-measured cholesterol withdrawal and predicted cholesterol withdrawal. As an example, Model 2 has the following values:
Model 2 - Predicted global cholesterol withdrawal = 5.27 + (230.53) ^* ApoC3 + (4786.52) ^* ApoE + (41.71) ^* ApoL1 + (14368.89) ^* PLTP + (-93.37) ^* SAA 1/2.

Logistic regression was then applied to this panel of proteins in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The model fitted to compare healthy controls and CAD patients (Model 3) had an area under the ROC curve (AUC) equal to 0.62 based on 5-fold cross-validation. The model fitted to compare healthy controls and CAD patients with an event (model 4) had an area under the ROC curve (AUC) equal to 0.64 based on a 5-fold cross-validation. The model fitted to compare healthy controls and event-free CAD patients (Model 5) had an area under the ROC curve (AUC) equal to 0.66 based on 5-fold cross-validation. The values for the three models are shown in Table 3A above and the equations are shown below.
CAD (with/without event) risk score = (i) + (c1) ^* ApoC3 + (c2) ^* ApoE + (c3) ^* ApoL1 + (c4) ^* PLTP + (c5) ^* SAA1/2
Model 3 - Total CAD Risk Score = 2.49 + (11.44) ^* ApoC3 + (-1434.48) ^* ApoE + (17.74) ^* ApoL1 + (-9141.23) ^* PLTP + (16.45) ^* SAA1/2 .
Model 4 - CAD risk score with event = 2.42 + (7.8) ^* ApoC3 + (-1597.95) ^* ApoE + (-28) ^* ApoL1 + (-9584.47) ^* PLTP + (12.03) ^* SAA1/ 2.
Model 5—No event CAD risk score = 1.08 + (13.1) ^* ApoC3 + (-1221.33) ^* ApoE + (73.58) ^* ApoL1 + (-8509.05) ^* PLTP + (24.59) ^* SAA1 /2.
Note that the values in Table 3A also include the number of 95% confidence interval ranges. Numbers in this range can be substituted into the formula for each value in the formula.

These formulas and values were then used to calculate final values for three patients (P1, P2, and P3). The results are shown in Table 3B below.

The risk score results in Table 3B can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)). In addition, this value can be multiplied by 100 to obtain a risk ratio.

For example, for patient P1, the risk score result for Model 3 in Table 3A is -1.16. Substituting this into the equation gives: 1/(1+exp(-(-1.16)))=0.2386673. Multiplying the result by 100 gives a risk probability of CAD for patient P1 of 23.8%. This same formula can be used for the results reported below for all of the models and patients in Table 3B, as well as the other panels in this example and Example 2 below.

Risk scores can be used by health care professionals to help determine treatment of patients with suspected heart disease or to exclude patients who may require treatment or monitoring. For example, in general, a patient with a risk rate of less than 10% for CAD is typically considered low risk (e.g., can be considered a patient who does not require treatment or monitoring for heart disease) and a risk rate for CAD of 10%. Patients at ~20% risk are usually considered intermediate risk (e.g., patients can be monitored for additional signs of heart disease), and patients with >20% risk of CAD are usually considered high risk. considered a risk (eg, can be treated immediately with therapeutic or surgical intervention).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 3C for Models 1, 3, 4, and 5. Since these coefficients are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 3C also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula. For example, the value of ApoC3 is not exactly 150.35, but can be anywhere between 109.91 and 190.8, which can be used in the model.

panel 2
Models 1-5 were generated to predict panel 2 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 2 are shown below:
Models 1-5 of panel 2 = (i) + (c1) ^* ApoC2 + (c2) ^* ApoC3 + (c3) ^* ApoE + (c4) ^* ApoL1 + (c5) ^* CLU + (c6) ^* PLTP
The values for Models 1-5 are shown in Table 4 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.77. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). The values for this model (Model 2) are shown in Table 4 above, and the correlation between laboratory measured and predicted cholesterol withdrawal was 0.6.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.61 based on 5-fold cross-validation. Models fitted to compare healthy controls and CAD patients with events had an area under the ROC curve (AUC) equal to 0.7 based on 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.59 based on 5-fold cross-validation. Values for the three models are shown in Table 4. Note that the values in Table 4 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 patients are shown in Table 5A below.

The results in Table 5A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 5b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 5B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 3
Models 1-5 were generated to predict panel 3 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 3 are shown below:
Models 1-5 of panel 3 = (i) + (c1) ^* ApoC2 + (c2) ^* ApoC3 + (c3) ^* ApoL1 + (c4) ^* PLTP
The values for Models 1-5 are shown in Table 6 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.73. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 6 and showed a correlation of 0.5 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.61 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.68 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.61 based on 5-fold cross-validation. Three CAD models are shown in Table 6. Note that the values in Table 6 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on the 5 models and 3 patients are shown in Table 7A below.

The results in Table 7A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 7b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 7B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 4
Models 1-5 were generated to predict panel 4 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 4 are shown below:
Models 1-5 of panel 4 = (i) + (c1) ^* ApoC1 + (c2) ^* ApoC3 + (c3) ^* CLU + (c4) ^* PLTP + (c5) ^* SAA4
The values for Models 1-5 are shown in Table 8 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.78. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). The values for this model are shown in Table 8 above, and the correlation between laboratory measured and predicted cholesterol withdrawal was 0.68.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.72 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with events had an area under the ROC curve (AUC) equal to 0.8 based on 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.66 based on a 5-fold cross-validation. The values for the three models are shown in Table 8 above. Note that the values in Table 4 also include numbers within the 95% confidence interval. For each value, numbers in this range can be substituted into the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 patients are shown in Table 9A.

The results in Table 9A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 9b for Models 1, 3, 4, and 5. Since these coefficients are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 9B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 5
Models 1-5 were generated to predict panel 5 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 5 are shown below:
Models 1-5 of panel 5 = (i) + (c1) ^* ApoA1 + (c2) ^* ApoC1 + (c3) ^* ApoC3
The values for Models 1-5 are shown in Table 10 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.85. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 10 above and had a correlation of 0.53 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.75 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.81 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.68 based on a 5-fold cross-validation. Table 10 shows the values for the three CAD models. Note that the values in Table 10 also include numbers within the 95% confidence interval. For each value, numbers in this range can be substituted into the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 patients are shown in Table 11A below.

The results in Table 11A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(-risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 11b for Models 1, 3, 4, and 5. Since these coefficients are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 11B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 6
Models 1-5 were generated to predict panel 6 overall cholesterol withdrawal and CAD risk. Models 1-5 of Panel 6 are shown below:
Models 1-5 of panel 6 = (i) + (c1) ^* ApoC3 + (c2) ^* ApoL1 + (c3) ^* PLTP + (c4) ^* SAA1/2
The values for Models 1-5 are shown in Table 12 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.74. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 12 and showed a correlation of 0.49 between laboratory measured and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.6 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with events had an area under the ROC curve (AUC) equal to 0.63 based on 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.64 based on a 5-fold cross-validation. Values for the three models (models 3-5) for CAD risk are shown in Table 12. Note that the values in Table 12 also include numbers within the 95% confidence interval. For each value, numbers in this range can be substituted into the formula.

The predicted cholesterol withdrawal and CAD risk based on the 5 models and 3 patients are shown in Table 13A below.

The results in Table 13A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(-risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 13b for Models 1, 3, 4, and 5. Since these coefficients are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 13B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 7
Models 1-5 were generated to predict panel 7 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 7 are shown below:
Models 1-5 of panel 7 = (i) + (c1) ^* ApoC1 + (c2) ^* ApoC3 + (c3) ^* ApoM
The values for Models 1-5 are shown in Table 14 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.84. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 14 and showed a correlation of 0.54 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.75 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.82 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.73 based on a 5-fold cross-validation. The values for the three models are shown in Table 14 above. Note that the values in Table 14 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 15A below.

The results in Table 15A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 15b for Models 1, 3, 4, and 5. Since these coefficients are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 15B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 8
Models 1-5 were generated to predict panel 8 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 8 are shown below:
Models 1-5 of panel 8 = (i) + (c1) ^* ApoC3 + (c2) ^* SAA1/2
The values for Models 1-5 are shown in Table 16 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.8. Linear regression was then applied to the proteins of this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 16 and showed a correlation of 0.48 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.58 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.54 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.53 based on a 5-fold cross-validation. The values for the three models are shown in Table 16 above. Note that the values in Table 16 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 17A below.

The results in Table 17A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 17b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 17B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 9
Models 1-5 were generated to predict panel 9 overall cholesterol withdrawal and CAD risk. Models 1-5 of Panel 9 are shown below:
Models 1-5 in panel 9 = (i) + (c1) ^* ApoC1 + (c2) ^* ApoC3
The values for Models 1-5 are shown in Table 18 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.85. Linear regression was then applied to the proteins of this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 18 and showed a correlation of 0.54 between laboratory measured and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.7 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.77 based on a 5-fold cross-validation. The model fitted to compare healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.68. The values for the three models are shown in Table 18 above. Note that the values in Table 18 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 19A below.

The results in Table 19A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 19b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 19B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 10
Models 1-5 were generated to predict panel 10's overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 10 are shown below:
Models 1-5 of panel 10 = (i) + (c1) ^* ApoC3
The values for Models 1-5 are shown in Table 20 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.8. Linear regression was then applied to the proteins of this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 20 and showed a correlation of 0.48 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.56 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.58 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.49 based on 5-fold cross-validation. The values for the three models are shown in Table 20 above. Note that the values in Table 20 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 21A below.

The results in Table 21A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)). This value can be multiplied by 100 to obtain the risk ratio for CAD.

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 21b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 21B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel eleven
Models 1-5 were generated to predict panel 11 overall cholesterol withdrawal and CAD risk. Models 1-5 of Panel 11 are shown below:
Models 1-5 of panel 11 = (i) + (c1) ^* ApoC3 + (c2) ^* ApoL1 + (c3) ^* SAA1/2
The values for Models 1-5 are shown in Table 22 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.8. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 22 and showed a correlation of 0.48 between laboratory measured and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.54 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.53 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.48 based on a 5-fold cross-validation. The values for the three models are shown in Table 22 above. Note that the values in Table 22 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 23A below.

The results in Table 23A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(-risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 23b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 23B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 12
Models 1-5 were generated to predict panel 12 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 12 are shown below:
Models 1-5 of panel 12 = (i) + (c1) ^* ApoA1 + (c2) ^* ApoC1 + (c3) ^* ApoC3 + (c4) ^* CLU
The values for Models 1-5 are shown in Table 24 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.86. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 24 and showed a correlation of 0.61 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.75 based on 5-fold cross-validation. The model fitted to compare healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.82, based on 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.63 based on 5-fold cross-validation. The values for the three models are shown in Table 24 above. Note that the values in Table 24 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 25A below.

The results in Table 25A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 25b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 25B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 13
Models 1-5 were generated to predict panel 13 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 13 are shown below:
Models 1-5 of panel 13 = (i) + (c1) ^* ApoC1 + (c2) ^* ApoC3 + (c3) ^* ApoL1 + (c4) ^* HP + (c5) ^* PLTP
The values for Models 1-5 are shown in Table 26 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.74. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 26 and showed a correlation of 0.54 between laboratory measured and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.7 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.79 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.68 based on a 5-fold cross-validation. The values for the three models are shown in Table 14 above. Note that the values in Table 26 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 27A below.

The results in Table 27A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 27b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 27B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 14
Models 1-5 were generated to predict panel 14 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 14 are shown below:
Models 1-5 of panel 14 = (i) + (c1) ^* ApoC3 + (c2) ^* ApoD + (c3) ^* SAA1/2
The values for Models 1-5 are shown in Table 28 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.8. Linear regression was then applied to the proteins of this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 28 and showed a correlation of 0.48 between laboratory measured and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.68 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.71 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.65 based on a 5-fold cross-validation. The values for the three models are shown in Table 28 above. Note that the values in Table 28 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 29A below.

The results in Table 29A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 29b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 29B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 15
Models 1-5 were generated to predict panel 15 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 15 are shown below:
Models 1-5 of panel 15 = (i) + (c1) ^* ApoA2 + (c2) ^* ApoC2 + (c3) ^* ApoC3 + (c4) ^* ApoD + (c5) ^* CLU + (c6) ^* SAA1/2 model 1-5 values are shown in Table 30 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.73. Partial linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 20 and showed a correlation of 0.57 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.61 based on 5-fold cross-validation. Models fitted to compare healthy controls and CAD patients with events had an area under the ROC curve (AUC) equal to 0.7 based on 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.6 based on 5-fold cross-validation. The values for the three models are shown in Table 30 above. Note that the values in Table 30 also include the number of 95% confidence interval ranges. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 31A below.

The results in Table 31A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 31b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 31B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 16
Models 1-5 were generated to predict panel 16 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 16 are shown below:
Models 1-5 of panel 16 = (i) + (c1) ^* ApoA2 + (c2) ^* ApoC2 + (c3) ^* ApoC3 + (c4) ^* ApoD + (c5) ^* ApoM + (c6) ^* SAA1/2
The values for Models 1-5 are shown in Table 32 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.65. Partial linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 32 and showed a correlation of 0.51 between laboratory measured and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.65 based on 5-fold cross-validation. The model fitted to compare healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.74 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.65 based on a 5-fold cross-validation. The values for the three CAD models are shown in Table 32 above. Note that the values in Table 32 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 33A below.

The results in Table 33A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 33b for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 33B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 17
Models 1-5 were generated to predict panel 17 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 17 are shown below:
Models 1-5 of panel 17 = (i) + (c1) ^* ApoC1 + (c2) ^* ApoC2 + (c3) ^* ApoC3
The values for Models 1-5 are shown in Table 34 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.81. An elastic net model was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 34 above and had a correlation of 0.53 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.74 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.78 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.72 based on 5-fold cross-validation. The values for the three models are shown in Table 34 above. Note that the values in Table 34 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 35A below.

The results in Table 35A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 35B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 35B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 18
Models 1-5 were generated to predict panel 18 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 18 are shown below:
Models 1-5 of panel 18 = (i) + (c1) ^* ApoA1 + (c2) ^* ApoC1 + (c3) ^* ApoC2 + (c4) ^* ApoC3 + (c5) ^* ApoC4
The values for Models 1-5 are shown in Table 36 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.86. Linear regression was then applied to the proteins of this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 36 and showed a correlation of 0.55 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.77 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with events had an area under the ROC curve (AUC) equal to 0.81 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.72 based on 5-fold cross-validation. The values for the three CAD models are shown in Table 36 above. Note that the values in Table 36 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 37A below.

The results in Table 37A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(-risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 37B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 37B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 19
Models 1-5 were generated to predict panel 19 overall cholesterol withdrawal and CAD risk. Models 1-5 of Panel 19 are shown below:
Models 1-5 of panel 19 = (i) + (c1) ^* ApoA2 + (c2) ^* ApoC1 + (c3) ^* ApoC2 + (c4) ^* ApoC3 + (c5) ^* ApoD + (c6) ^* SAA1/2
The values for Models 1-5 are shown in Table 38 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.71. Partial linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). The values for this model (Model 2) are shown in Table 38, with a correlation of 0.51 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.74 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.79 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.73 based on a 5-fold cross-validation. The values for the three CAD models are shown in Table 38 above. Note that the values in Table 38 also include the number of 95% confidence interval ranges. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 39A below.

The results in Table 39A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(-risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 39B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 39B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 20
Models 1-5 were generated to predict panel 20's overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 20 are shown below:
Models 1-5 of panel 20 = (i) + (c1) ^* ApoA2 + (c2) ^* ApoC3 + (c3) ^* ApoD + (c4) ^* ApoE + (c5) ^* ApoL1 + (c6) ^* PLTP + (c7) ^* SAA1/2
The values for Models 1-5 of panel 20 are shown in Table 40 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.8. An elastic net model was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 40 above and had a correlation of 0.52 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.64 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.68 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.66 based on a 5-fold cross-validation. The values for the three CAD models are shown in Table 40 above. Note that the values in Table 40 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 41 below.

The results in Table 41A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 41B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 41B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 21
Models 1-5 were generated to predict panel 21 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 21 are shown below:
Models 1-5 of panel 21 = (i) + (c1) ^* ApoC3 + (c2) ^* ApoM + (c3) ^* PLTP + (c4) ^* SAA1/2
The values for Models 1-5 of Panel 21 are shown in Table 42 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.76. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 42 above and had a correlation of 0.5 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.65 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.71 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.68 based on a 5-fold cross-validation. The values for the three CAD models are shown in Table 42 above. Note that the values in Table 42 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 43A below.

The results in Table 43A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(-risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 43B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 43B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 22
Models 1-5 were generated to predict panel 22 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 22 are shown below:
Models 1-5 of panel 22 = (i) + (c1) ^* ApoC3 + (c2) ^* ApoD + (c3) ^* PLTP + (c4) ^* SAA1/2
The values for Models 1-5 are shown in Table 44 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.79. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 44 above and had a correlation of 0.5 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.65 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.71 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.66 based on a 5-fold cross-validation. The values for the three CAD models are shown in Table 44 above. Note that the values in Table 44 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 45A below.

The results in Table 45A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 45B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 45B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 23
Models 1-5 were generated to predict panel 23's overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 23 are shown below:
Models 1-5 of panel 23 = (i) + (c1) ^* ApoA2 + (c2) ^* ApoC3 + (c3) ^* ApoD + (c4) ^* ApoL1 + (c5) ^* ApoM + (c6) ^* PLTP + (c7) ^* SAA1/2
The values for Models 1-5 are shown in Table 46 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.79. An elastic net model was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 46 above and had a correlation of 0.5 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.65 based on 5-fold cross-validation. Models fitted to compare healthy controls and CAD patients with events had an area under the ROC curve (AUC) equal to 0.7 based on 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.67 based on a 5-fold cross-validation. The values for the three CAD models are shown in Table 46 above. Note that the values in Table 46 also include the number of 95% confidence interval ranges. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 47A below.

The results in Table 47A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 47B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 47B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 24
Models 1-5 were generated to predict panel 24 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 24 are shown below:
Models 1-5 of panel 24 = (i) + (c1) ^* ApoC1 + (c2) ^* ApoC3 + (c3) ^* SAA1/2
The values for Models 1-5 are shown in Table 48 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.85. Linear regression was then applied to the proteins of this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 48 above and had a correlation of 0.53 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.7 based on 5-fold cross-validation. The model fitted to compare healthy controls and CAD patients with events had an area under the ROC curve (AUC) equal to 0.75 based on 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.67 based on a 5-fold cross-validation. The values for the three CAD models are shown in Table 48 above. Note that the values in Table 48 also include the number of 95% confidence interval ranges. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 49A below.

The results in Table 49A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(-risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 49B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 49B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 25
Models 1-5 were generated to predict panel 25 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 25 are shown below:
Models 1-5 of panel 25 = (i) + (c1) ^* ApoC1 + (c2) ^* ApoC3 + (c3) ^* C3 + (c4) ^* PLTP
The values for Models 1-5 are shown in Table 50 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.79. Robust linear regression was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 50 and had a correlation of 0.56 between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.74 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with events had an area under the ROC curve (AUC) equal to 0.8 based on 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.7 based on 5-fold cross-validation. The values for the three CAD models are shown in Table 50 above. Note that the values in Table 50 also include the number of 95% confidence interval ranges. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 51A below.

The results in Table 51A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(-risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 51B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 51B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 26
Models 1-5 were generated to predict panel 26 overall cholesterol withdrawal and CAD risk. Models 1-5 of panel 26 are shown below:
Models 1-5 of panel 26 = (i) + (c1) ^* ApoA2 + (c2) ^* ApoC2 + (c3) ^* ApoC3 + (c4) ^* ApoD + (c5) ^* ApoE + (c6) ^* ApoL1 + (c7) ^* ApoM + (c8) ^* CETP+(c9) ^* CLU+(c10) ^* PLTP+(c11) ^* PON1+(c12) ^* SAA1/2
The values for Models 1-5 are shown in Table 52 below.

This model was then tested on 35 replicate samples. The correlation between laboratory measured cholesterol withdrawal and predicted cholesterol withdrawal was 0.78. An elastic net model was then applied to the proteins in this panel in another cohort of samples consisting of 76 fresh and frozen samples (40 fresh and 36 frozen). This model (Model 2) is shown in Table 52 above, and the correlation between laboratory measured and predicted cholesterol withdrawal was 0.62.

Finally, logistic regression was applied to the proteins of this panel in another cohort of samples composed of 74 healthy controls and 157 CAD patients (83 with events and 74 without events). The fitted model comparing healthy controls and CAD patients had an area under the ROC curve (AUC) equal to 0.63 based on 5-fold cross-validation. The fitted model comparing healthy controls and CAD patients with an event had an area under the ROC curve (AUC) equal to 0.71 based on a 5-fold cross-validation. The fitted model comparing healthy controls and event-free CAD patients had an area under the ROC curve (AUC) equal to 0.6 based on 5-fold cross-validation. The values for the three CAD models are shown in Table 52 above. Note that the values in Table 52 also include numbers within the 95% confidence interval. Numbers in this range can be substituted into the formula for each value in the formula.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 53A below.

The results in Table 53A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As above, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 53B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 53B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

Example 2
In this example, different algorithms for ABCA1 cholesterol abstraction and coronary artery disease risk were used, and global ABCA1 cholesterol abstraction (not total cholesterol abstraction) and total ABCA1 cholesterol abstraction were analyzed by five panels of single and multiple HDL proteins. We describe testing our ability to predict cardiovascular disease risk. For the five panels, the same assays and samples were measured as in Example 1 above. The five panels are panels 27-30 and panel 10, as shown in Table 54 below.

panel 27
Models 1-5 were generated to predict panel 27 ABCA1 cholesterol abstraction and CAD risk. Models 1-5 of panel 27 are shown below:
Models 1-5 in panel 27 = predicted ABCA1 cholesterol abstraction = (i) + (c1) ^* ApoC3 + (c2) ^* ApoE + (c3) ^* ApoL1 + (c4) ^* HP + (c5) ^* PLTP
The values for Models 1-5 are shown in Table 55 below.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 56A below.

The results in Table 56A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As described above in Example 1, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 56B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 56B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 28
Models 1-5 were generated to predict panel 28 ABCA1 cholesterol abstraction and CAD risk. Models 1-5 of panel 28 are shown below:
Models 1-5 in panel 28 = predicted ABCA1 cholesterol abstraction = (i) + (c1) ^* ApoA1 + (c2) ^* ApoA2 + (c3) ^* ApoC1 + (c4) ^* ApoC2 + (c5) ^* ApoC3 + (c6) ^* ApoC4 + ( c7) ^* ApoD + (c8) ^* ApoE + (c9) ^* ApoL1 + (c10) ^* ApoM + (c11) ^* C3 + (c12) ^* CLU + (c13) ^* HP + (c14) ^* SAA1/2 + (c15) ^* SAA4
The values for Models 1-5 are shown in Table 57 below.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 58A below.

The results in Table 58A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As described above in Example 1, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 58B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 58B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 29
Models 1-5 were generated to predict panel 29 ABCA1 cholesterol abstraction and CAD risk. Models 1-5 of panel 29 are shown below:
Models 1-5 in panel 29 = predicted ABCA1 cholesterol abstraction = (i) + (c1) ^* ApoA1 + (c2) ^* ApoC3 + (c3) ^* ApoD + (c4) ^* ApoE + (c5) ^* ApoL1 + (c6) ^* ApoM + ( c7) ^* HP + (c8) ^* PLTP + (c9) ^* PON1 + (c10) ^* SAA1/2
The values for Models 1-5 are shown in Table 59 below.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 60A below.

The results in Table 60A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As described above in Example 1, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 60B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 60B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 30
Models 1-5 were generated to predict ABCA1 cholesterol abstraction and CAD risk in panel 30. Models 1-5 of panel 30 are shown below:
Models 1-5 in panel 30 = predicted ABCA1 cholesterol abstraction = (i) + (c1) ^* ApoA1 + (c2) ^* ApoA2 + (c3) ^* ApoC3 + (c4) ^* ApoC4 + (c5) ^* ApoD + (c6) ^* ApoE + ( c7) ^* ApoL1 + (c8) ^* ApoM + (c9) ^* C3 + (c10) ^* HP + (c11) ^* PLTP + (c12) ^* PON1 + (c13) ^* SAA1/2
The values for Models 1-5 are shown in Table 61 below.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 62A below.

The results in Table 62A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As described above in Example 1, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 62B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 62B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

panel 10
Models 1-5 were generated to predict panel 10 ABCA1 cholesterol withdrawal and CAD risk. Models 1-5 of panel 10 are shown below:
Panel 10 Models 1-5 - predicted ABCA1 cholesterol abstraction = (i) + (c1) ^* ApoC3
The values for Models 1-5 are shown in Table 63 below.

The predicted cholesterol withdrawal and CAD risks calculated based on 5 models and 3 specific patients are shown in Table 64A below.

The results in Table 64A can be used to obtain the probability of CAD based on the following formula: probability = 1/(1 + exp(- risk score)).

As described above in Example 1, absolute molar amounts of each protein were determined using absolute amounts of internal standards. These molar amounts were then used to determine the coefficients in Table 64B for Models 1, 3, 4, and 5. Since these factors are in units of molar mass, they can be used in models where protein values are determined by any kind of detection assay (eg, ELISA, mass spectroscopy, etc.).

Note that the values in Table 64B also include the number of 95% confidence interval ranges. For each value in the formula (model), numbers in this range can be substituted into the formula.

Example 3
Apolipoprotein A-I Associated Proteome Associated with Cholesterol Extraction Ability and Coronary Artery Disease In this example, the ApoA-I associated serum proteome and its relationship to HDL function in relation to cholesterol extraction were discussed. Furthermore, in this study, an attempt was made to construct a prediction method for cholesterol extraction capacity (CEC). Serum protein fractions associated with His-tagged recombinant ApoA-I were subjected to data-dependent proteomic analysis to develop a targeted quantitative proteomic method for 21 proteins associated with reverse cholesterol transport and/or coronary artery disease (CAD). This targeting method was compared to cell-based CEC measurements (N=105) to derive a proteome prediction algorithm that was applied to healthy subjects (N=74) as well as CAD subjects with major adverse cardiovascular events (N=105). =83), and a case/control study of CAD specimens without major adverse cardiovascular events (N=74). There were significant parallels between the ApoA-I-associated serum proteome and proteins found in HDL. A proteome prediction algorithm composed of ApoA-II, ApoC-I, ApoC-II, ApoC-III, ApoD, and serum amyloid A (panel 19) correlated strongly with cell-based CEC assay results (R=0. 77). Proteome-predicted CEC measurements in the case/control cohort showed a significant inverse correlation between CEC and CAD diagnosis (P=0.0032), and a significant CEC between control and CAD specimens without MACE. A reduction was observed (P=0.04), with a further decrease in CEC observed in CAD specimens with MACE.

Materials and Methods All study methods were approved by local institutional review boards as required. All reagents were purchased from Sigma-Aldrich (St. Louis, Mo.) or the highest grade available from ThermoFisher unless otherwise stated. Specimens used in this study were obtained from Golden West biologicals, which supplies pooled serum samples. Lipoprotein-free serum prepared by immunoaffinity depletion was purchased from GenwayBio (San Diego, Calif.). In addition, remnant samples obtained from Cleveland HeartLab were used. Clinical specimens representing cases of coronary artery disease and controls were obtained from the Fairbanks Institute for Healthy Communities.

Enrichment of ApoA-I-Related Serum Fractions Recombinantly expressed and purified 0.5 mg/mL ¹⁵ N-labeled His ₆ -tagged apolipoprotein AI ( ¹⁵ N-His ₆ ApoA-I) (Genscript, Piscataway, NJ) of 1×PBS (pH 7.4) solution was added to 12 μL of human serum. The ¹⁵ N-His ₆ ApoA-I/serum mixture was incubated at 37° C. for 30 minutes. After incubation, the samples were diluted to a total volume of 200 μL with 5 mm imidazole, 20 mM sodium phosphate, 150 mM sodium chloride, pH 8.0, according to the manufacturer's protocol, and then loaded with 5 μL of Ni-NTA HisBind Superflow stationary phase, PhyTip (Phynexus Purification was carried out using a Tecan Freedom Evo automated liquid handler (Tecan, Mannedorf, Switzerland) equipped with a Tecan, San Jose, Calif.). Briefly, the diluted sample was slowly bound to the Phytip column at 250 μL/min using repeated pipette cycles, followed by subsequent washes with 300 μL of 20 mM imidazole, 20 mM sodium phosphate, 150 mM sodium chloride, pH 8.0. . Bound His ₆ -ApoA-I and related species were then eluted with 300 mM imidazole, 50 mM Tris-HCl, pH 9.0, 25% methanol. After elution, samples were either used immediately for subsequent analysis or stored at -80°C until required.

Data-dependent proteomic analysis by nanoLC-MS In addition to 500 ng of endoproteinase LysC (Wako Chemicals USA, Richmond, VA), dithiothreitol was added to the concentrated ApoA-I related serum fractions to a final concentration of 5 mM. Samples were digested in an Eppendorf PCR thermal cycler at 37°C for 4 hours and then kept at 4°C to stop endoproteinase activity. The resulting peptide mixture was separated using a Waters nanoACQUITY Ultra High Pressure LC system (Waters, Inc., Milford, Mass.). A 10 μL volume containing 500 ng of total peptide material was loaded onto a reverse-phase Symmetry C18 trapping column (180 μm id x 20 mm, 5 μm particles) (Waters, Inc) using mobile phase A (0.1% formic acid, 2% acetonitrile in water). Inject, capture and wash at 10 μL/min for 4 minutes. Peptides were then eluted at 300 nL/min on a nanoACQUITY column (75 μm id×250 mm) packed with 1.7 μm BEH130 C18 stationary phase. In this case, mobile phase B (0.1% formic acid, 2% acetonitrile in water) was used with a gradient of 2% to 50% for 210 minutes, followed by increasing mobile phase B to 100% for 15 minutes, followed by 2% Re-equilibrate with mobile phase B for 15 minutes. LTQ-Orbitrap using precursor ion scans from m/z 350-1800 in a 120,000 resolution Orbitrap followed by ion trapping using collision-induced dissociation for 25 data-dependent MS/MS events Peptides were detected on a Velos mass spectrometer (Thermo Fisher Scientific, Waltham, Mass.). Precursors were separated with a selection width of 2.5 m/z and fragmented at 35% normalized collision energy for 20 ms. Monoisotopic precursor selection was possible and precursors with 1+ and empty charge states were excluded from MS/MS selection. Dynamic exclusion of precursors with mass exclusion windows of plus or minus 10 ppm for 60 seconds was utilized. The automatic gain control limits for the ion trap and orbitrap were set to 1×10 ⁴ and 1×10 ⁶ respectively. Mass spectrometry data files in RAW format were processed with MaxQuant (version 1.4.1.2) ¹⁴ using the integrated Andromeda search algorithm ¹⁵ to identify proteins. In addition to the Common Pollutant Database, searches were performed against the UniProt Human Database (downloaded from the European Bioinformatics Institute on 14 October 2013, consisting of 88,304 entries). The search was performed with a precursor mass tolerance of ±7 ppm and a fragment mass tolerance of 0.6 Da. LysC was set as the cleaving enzyme and peptides containing at least 6 amino acids and up to 2 cleavage errors were analyzed. Methionine oxidation was allowed as a variable modification. The false positive rate for proteins and peptides was set at 1%. The protein identification was further refined so that at least two peptides were identified.

Targeted proteome analysis Reducing agent (5 μL of 100 mM dithiothreitol) and protease (10 μL of 50 ng/μL endoproteinase LysC (Wako)) were added to 85 μL of concentrated ApoA-I related serum fraction and incubated at 37° C. for 4 hours. . The temperature at that point had dropped to 4°C. Five μL of a defined mixture of ¹³ C ₆ , ¹⁵ N ₂ -lysine-labeled internal standard peptides was added to 75 μL of protein digest, followed by a 25 μL injection and analysis by liquid chromatography multiple reaction monitoring mass spectrometry (LC-MRM). The injected sample was packed and washed on the column for 1.25 minutes before being eluted with a linear gradient of mobile phase B at 500 μL/min. Peptides were detected using an Agilent 6490 triple quadrupole mass spectrometer operating in dynamic MRM mode, which allows targeted detection of peptide targets within scheduled retention windows. Transitions are selected, optimized and determined to be specific to the target peptide within the sample. Two transitions were monitored per peptide and a maximum of two peptides per protein. A detailed list of peptide targets and their transitions is provided in Table 1 above and for Panel 19 in Table 67 below.

Peptide signal intensities were obtained by integrating chromatographic peaks against quantifier transitions using MassHunter Quantitative Analysis software (Agilent). All peaks were manually reviewed using qualifier ion ratios and internal standard peaks.

Isolation of HDL by Ultracentrifugation HDL from pooled serum (1.063 g/mL<p<1.21 g/mL) was isolated using the method of Brewer with minor modifications ¹⁶ .

Cell-Based Evaluation of Cholesterol Extraction Ability Human serum samples were LDL-depleted and a cell-based assay measuring ³ H-labeled cholesterol withdrawal from J774 macrophages was performed using the method described by de la Llera-Moya ¹⁷ using Vascular Strategies, Inc. , (Plymouth Meeting, Pa.) was outsourced. All measurements were reported normalized to the pull-out value.

CEC Algorithm Development Specimens Serum samples for development of the pull-out correlation model were from CHL unidentified remnant specimens collected in two batches, 6 weeks apart, to provide training and test sets, respectively. Taken. Quantitative analysis of LDL-c, HDL-c, ApoA, ApoB, triglycerides, and high-sensitivity C-reactive protein (hsCRP) was used to guide the selection of candidate samples for each set, ultimately resulting in a set of suitable specimens. (Table 65).

Quantitative ApoA-I-associated proteomic analysis and CEC measurements were collected on 105 specimens with highly characterized serum pools as quality control material.

Clinical Confirmation Specimens Specimens were selected from the Fairbanks Institute for Healthy Communities biobank. It consisted of serum samples from 1500 men and women aged 22-87 years, including 750 with coronary angiographic findings of CAD (≥50% occlusion), CAD, positive stress test, diabetes, hypertension, or dyslipidemia. There were 750 control subjects with no positive findings for (LDL-C≧130 mg/dL, HDL-C≦40 mg/dL, total cholesterol≧240 mg/dL, or triglycerides≧200 mg/dL). Fasting blood samples were collected according to the test SOP and then stored at -80°C. Subjects diagnosed with CAD were evaluated to determine two groups: cases and cases with events. All CAD patients were treated for myocardial infarction (410), coronary artery bypass or angioplasty (36.1, 45.82), or stroke (434. 91). If myocardial infarction was confirmed by ICD9 classification, records were reviewed and patients with 2 out of 3 history of ischemic pain, abnormal ECG, abnormal troponin were selected. In total, 74 CAD subjects without events and 83 CAD subjects with MACE events were selected (Table 66).

A matched control group of 74 individuals was also selected.

Target Algorithm Development Prior to computational analysis, the peak area abundance of each target peptide was normalized by the intensity of ¹⁵ N-His ₆ ApoA-I used in the enrichment process. For proteins with more than one peptide measured by LC-MS/MS, the peptide showing the highest intensity in the majority of the samples analyzed was used to establish the relative amount of protein. Protein level data were used for statistical analysis. An analytical pipeline involving a series of sequential steps for feature selection was applied to the normalized protein level data to search for proteins associated with CEC ¹⁸ . For univariate analysis, robust linear regression was applied to each protein to predict CEC on 70 training samples, and proteins with p-value < 0.1 were selected. Further filtering was applied using the elastic net model. For multivariate analysis, biomarker panels were constructed with proteins filtered using partial least squares regression. To assess the performance of the biomarker panel, the Spearman correlation and median absolute deviation (cost) between predicted and measured CEC were calculated. First, a biomarker panel was generated using uncalibrated LC-MS peak intensities. The proteomic method was later refined to provide post-calibration absolute molar amounts, and panel coefficients were refined with post-calibration MRM data using partial least squares regression. Finally, the panel was tested with 35 validation samples. To assess whether the CEC predictive biomarker panel can distinguish subjects with and without CAD, we tested biomarker proteins in 74 healthy controls and 157 CAD patients. Statistical analysis of the data was performed using R version 3.2.3 (R Core Team (2013). R: A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria). Comparison of patient cohort cumulative distributions was performed using the Kolmogorov-Smirnov (KS) test.

result:
Rapid Preparation of ApoA-I Affinity Pools To facilitate accurate and high-throughput analysis of lipid-free ApoA-I-related proteins using mass spectrometry-based proteomics, a strategy for rapid affinity enrichment was devised. . This strategy used His-tagged ApoA-I and immobilized metal affinity chromatography (Fig. 2). Recombinant lipid-free ApoA-I with an N-terminal polyhistidine tag (His ₆ ApoA-I) was added to human serum and incubated at 37°C. The sample was then passed over a nickel-NTA stationary phase to allow binding of His ₆ ApoA-I and related proteins. After washing the stationary phase with low concentrations of imidazole to suppress non-specifically bound proteins, complexed proteins and excess His ₆ ApoA-I are eluted into the desired recovery buffer containing high concentrations of imidazole. . The eluted samples were then subjected to a standard enzymatic digestion workflow for final analysis by liquid chromatography-mass spectrometry (LC-MS/MS). ¹⁵ N-labeled His ₆ ApoA-I ( ¹⁵ N-His ₆ ApoA-I), which allows specific measurement of exogenous ApoA-I, was used to distinguish it from endogenous proteins belonging to the analytes to be analyzed.

Data-dependent proteomic analysis of lipid-free ApoA-I affinity pools As revealed in early data-dependent experiments, the majority of proteins identified in the affinity fraction of lipid-free ApoA-I are HDL-associated. I know An attempt was made to further clarify this observation by a qualitative proteomic comparison of HDL isolated using conventional ultracentrifugation and ApoA-I-related serum fractions using a commercially available serum pool of 400 male donors. rice field. Three replicate steps were performed for each enrichment technique. Overall, 91 proteins were identified in HDL purified by ultracentrifugation and 162 proteins were identified in the lipid-free ApoA-I affinity pool, of which 78 proteins were common to both preparation methods. (see Figure 3).

Proteins identified in our experiments were compared to the HDL Proteome Watch list, which summarizes the results of 17 published HDL proteome studies. This list contains 229 proteins, 95 of which have been identified in 3 or more studies or validated by other molecular biology techniques. Of the total 229 proteins on the Proteome Watch list, 80 were identified in HDL prepared by ultracentrifugation, 86 were identified in our affinity-purified pools, and 72 (minus common contaminants such as keratin). ) were common for both preparation methods (77% common) (Fig. 3).

Target Quantification by Multiple Reaction Monitoring Mass Spectrometry To investigate the relationship between ApoA-I affinity pools and cholesterol efflux capacity (CEC), we attempted to use an accurate quantification method based on liquid chromatography multiple reaction monitoring (LC-MRM). Based on a literature search and our data-dependent LC-MS results, 21 proteins known to be associated with lipid transport, reverse cholesterol transport and/or cardiovascular disease were selected for development (lipid metabolism ( Apolipoproteins A-I, A-II, A-IV, C-I, C-II, C-III, C-IV, D, E, F, J, LI, M), enzymes (phospholipid transport protein-PLTP, cholesteryl ester transfer protein-CETP, lecithin cholesterol acyltransferase-LCAT, paraoxonase 1-PON1), and acute phase response proteins (complement C3, haptoglobin, serum amyloid A1 and 2-SAA1/2, and SAA4 )). Two peptides (where possible) from each protein for assay development were identified and the overall workflow optimized. Digestion conditions were obtained that produced stable peptide abundances for all proteins within 4 hours. Recovery of native ApoA-I from serum stocks was estimated to be approximately 55%, which correlated strongly with ApoA-I levels in serum determined by immunoassay (Pearson r=0.84 , Supplementary Fig. 6). By comparison, the mean recovery of lipid-free ¹⁵ N-His ₆ ApoA-I was 85% from PBS buffer and 72% from solvent-depleted serum.

Development and Validation of a Multivariate Algorithm for CEC Prediction A lipid-free ApoA-I affinity enrichment technique was used to analyze a series of serum samples to determine if there is an association between affinity enriched proteins and CEC. examined. A set of 70 training test samples and 35 independent test samples (Table 65) were randomly selected regardless of any disease diagnosis, but with caution regarding concordance with lipoprotein measurements. Normalized CEC and normalized mass spectrometry data of the complement of 21 proteins were determined for each specimen. A multi-step informatics pipeline including univariate and multivariate statistics was performed to develop a biomarker panel whose relative protein abundance correlated with CEC ¹⁸ in cell lines. After univariate analysis, 9 proteins (apolipoproteins A-I, A-II, C-I, C-II, C-III, C-IV, D, CETP, and SAA) were identified by robust linear regression. In a subsequent elastic net regression, 6 proteins (ApoA-II, ApoC-I, ApoC-II, ApoC-III, ApoD, and SAA) were selected and subjected to partial linear regression to generate the final predictive model Confirmed.

The prediction panel performed well considering the measurement error of the CEC measurements (training set, Spearman r=0.63, p<0.001; validation set, Spearman r=0.70, P<0.001). . The performance of the proteome prediction CEC algorithm within the entire development sample set is shown in Figure 4A (Pearson r = 0.77). Associations between CECs determined by cell-based assays and proteomic estimates and other clinical measurements (total cholesterol, HDL-c, LDL-c, non-HDL-c, triglycerides, ApoA-I, ApoB, and hsCRP) examined. The most significant associations with cell line CEC were ^ApoA -I (Pearson r=0.57, FIG. 4B) and HDL-c (Pearson r=0.45, FIG. 4C). A significant negative correlation was also observed between CEC and hsCRP (Pearson r=−0.23, FIG. 4D). Predicted CEC behaved similarly for total cholesterol, but the relationship with ApoA-I and HDL-c was not calculated since ApoA-I was included in the proteome panel.

Predicted CEC is inversely correlated with CAD Given that predicted CEC should be inversely correlated with cardiovascular disease, similar to reports on cell-based CEC measurements, using our multivariate panel, 157 CAD patients (74 without major adverse cardiovascular events (MACE), 83 with MACE), and age- and sex-matched appearance predicted CEC for 74 healthy controls (Table 66).

The distribution of predicted CECs among these cohorts is shown in Figure 5A. CAD patients were found to have a lower median predicted CEC (9.54% withdrawal/4 hours; controls 9.98% withdrawal/4 hours; p=0.0032). When CAD cases are divided into MACE-free and non-specific adverse event cases, significantly lower predicted CEC is observed in MACE-free CAD patients compared to controls (P=0.04). CAD patients who underwent revascularization (n=29) had a lower predicted CEC than CAD patients without MACE (P=0.027, P=0.0003 compared to controls). CAD patients who experienced myocardial infarction (n=38) or stroke (n=16) showed a lower median predicted CEC, which was not significantly different from CAD patients without MACE. Comparisons of age- and sex-matched controls with CAD samples were repeated after 6 weeks to demonstrate support for trends observed in the initial analysis. A comparison of the performance of predictive CEC and cell-based CEC measurements was also performed in a small number of selected samples with clinical outcomes. Cell line CEC measurements were obtained after identifying the 15 highest and lowest predicted CEC values in the control and CAD cohorts, respectively. As shown in Supplementary Fig. 7, a subset of CAD specimens had significantly lower CECs (P-value < 0.001) as determined by both proteome prediction and cell-based CEC measurements. Furthermore, there was no apparent significant difference when comparing predictive values with cell line measurements in the same subset of outcomes (P=0.39 for control, P=0.08 for CAD).

Determination of CEC by cell-based assays has been shown to be effective in investigating cholesterol metabolism, and an increasing number of studies have shown that serum-based CEC measurements are independent predictors of cardiovascular risk. Unfortunately, this method is labor intensive, workflow complex, and relatively imprecise, making it unsuitable for clinical testing that requires scale, throughput, accuracy, calibration, and ultimately analytical validation. There are obstacles to diverting technology.

Performance of proteome-predicting CECs The performance of proteome-predicting CECs on the test set (r=0.71) is encouraging, suggesting an important role for proteins other than ApoA-I in regulating cholesterol efflux. A recent clinical study of CEC performing a modified cell-based assay using fluorescently labeled cholesterol found significant predictive value that was somewhat consistent with normalized withdrawal using the underlying radioisotope assay (R = 0.54) ¹² . In addition, a correlation with withdrawal has been confirmed. As expected, many studies reported positive univariate correlations with ApoA-I ranging from r = 0.22 to 0.64 (r = 0.57 in the present study) ^11,19 . The sphingomyelin/phosphoethanolamine ratio is significantly associated with withdrawal (r=0.64) in high HDL-C (>68 mg/dL) but not normal HDL-C levels. It has been proven. Retrieval is also known to be HDL subclass dependent, further suggesting a complex set of structural and functional interactions ²⁰ . Although these variations remain to be fully elucidated, additional biochemical and biophysical data may be integrated into the model to improve CEC prediction.

Mechanistic Considerations of CEC Prediction Panel Proteins The CEC proteome predictors obtained in our study are proteins mechanistically associated with cholesterol transport (apolipoproteins A-II, C-I, C-II, C- III, D, and SAA; panel 19). Among the proteins positively correlated with withdrawal, ApoA-I and ApoA-II function as primary receptors for cellular cholesterol ^1,21 . Similarly, ApoC-I also activates the lecithin-cholesterol acyltransferase LCAT, a minor but efficient cholesterol receptor, which plays a role in HDL maturation ²² . ApoC-II and ApoC-III regulate lipoprotein lipases involved in HDL remodeling, and ApoD plays a role in the formation of pre-β ₃ -HDL, a potent cellular cholesterol-receptive form of HDL ^22,23 . Serum amyloid A (SAA) is an acute phase reactive protein and has been extensively characterized to negatively affect cholesterol efflux capacity by displacement of ApoA-I from HDL ^24,25 . We also noted that the functional role of SAA in regulating withdrawal is reflected in the negative correlation between withdrawal and hsCRP. hsCRP is another commonly used biomarker for measuring systemic inflammation (although the present study did not identify relevance for ApoA-I). Interestingly, the interaction of lipid-free ApoA-I with serum proteins elucidated by proteomic analysis shows a high degree of overlap with the canonical HDL proteome classified in the Proteome Watch list.
References for this example:
1. Kontush A, Chapman MJ. High-Density Lipoproteins: Structure, Metabolism, Function, and Therapeutics. John Wiley & Sons, Inc;
2. Drew et al. , High-density lipoprotein and apolipoprotein AI increase endothelial NO synthesis activity by protein association and multisite phosphorylation. Proc
Natl Acad Sci. 2004; 101(18):6999-7004.
3. Ansell et al. Inflammatory/Antiinflammatory Properties of High-Density Lipoprotein Distinguish Patients from Control Subjects Better Than High-Density Lipoprotein Cholesterol Levels Favorable
Affected by Simvastatin Treatment. Circulation. 2003; 108(22):2751-2756.
4. Gaidukov L, Tawfik DS. High Affinity, Stability, and Lactonase Activity of Serum Paraoxonase PON1 Anchored on HDL with ApoA-I. Biochemistry. 2005;44(35):11843-11854.
5. Huang et al. Myeloperoxidase, paraoxonase-1, and HDL form a functional ternary complex. J Clin Invest. 2013; 123(9):3815-3828.
6. Cavigiolio et al. , Exchange of apolipoprotein AI between lipid-associated and lipid-free states: a potential target for oxidative generation of dysfunctional high density lipoprotein. J Biol Chem. 2010;285(24):18847-18857.
7. Fredolini et al. Immunocapture strategies in translational proteomics. Expert Rev Proteomics. 2016; 13(1):83-98.
8. Gavin et al. Functional organization of
the yeast proteome by systematic analysis of protein complexes. Nature. 2002;415(6868):141-147.
9. Gundry et al. , Investigation of an albumin-enriched fraction of human serum and
its albuminome. Proteomics-Clin Appl. 2007; 1(1):73-88.
10. Poulsen et al. , Serum Amyloid P Component (SAP) Interactome in Human Plasma Containing Physiological Calcium Levels. Biochemistry. 2017;56(6):896-902.
11. Khera, et al. Cholesterol efflux capacity, high-density lipoprotein function, and
atherosclerosis. N Engl J Med. 2011;364(2):127-135.
12. Rohatgi et al. HDL cholesterol efflux capacity and incident cardiovascular events. N Engl J Med. 2014;317(25):2383-2393.13. Mody et al. , Beyond Coronary Calculation, Family History, and C-Reactive Protein. J Am Coll Cardiol. 2016;67(21):2480-2487.
14. Cox J, Mann M.; MaxQuant enables high peptide identification rates, individualized p. p. b. -range mass accuracies and proteome-wide protein quantification. Nat Biotechnol. 2008;26(12):1367-1372.
15. Cox et al. , a peptide search engine integrated into the Max Quant environment. J.
Proteome Res. 2011; 10(4): 1794-1805.
16. Duverger et al. , Biochemical characterization of the three major subclasses of
lipoprotein AI preparatively isolated from human plasma. Biochemistry. 1993;32(46):12372-12379.
17. de la Llera-Moya et al. , The ability to promote efflux via ABCA1 determines the capacity of serum specifications with similar high-density lipoprotein cholesterol to remove cholesterol from cholesterol. Arterioscler Thromb Vasc Biol. 2010;30(4):796-801.
18. Sin et al. , Biomarker Development for Chronic Obstructive Pulmonary Disease. From Discovery to Clinical Implementation. Am J Respir Crit Care Med. 2015; 192(10): 1162-1170.
19. Saleheen et al. Association of HDL cholesterol efflux capacity with incident coronary heart disease events: a prospective case-control study. Lancet Diabetes &
Endocrinol. 2015;3(7):507-513.
20. Du et al. HDL particle size is a critical determinant of ABCA1-mediated macrophage cellular cholesterol export. Circ Res. 2015; 116(7): 1133-1142.
21. Rotllan et al. , Overexpression of Human Apolipoprotein A-II in Transgenic Mice
Does Not Impair Macrophage-Specific Reverse Cholesterol Transport In Vivo. Arterioscler Thromb Vasc Biol. 2005;25(9).
22. von Eckardstein A, Nofer JR, Assmann G.; High Density Lipoproteins and Arteriosclerosis: Role of Cholesterol Efflux and Reverse Cholesterol Transport. Arterioscler Thromb Vasc Biol. 2001;21(1):13-27.
23. Francone OL, Fielding CJ. Initial steps
in reverse cholesterol transport: the role of short-lived cholesterol acceptors. Eur HeartJ. 1990; 11 (suppl E).
24. Vaisar et al. Inflammatory remodeling of the HDL proteome impairments cholesterol efflux capacity. J Lipid Res. 2015;56(8):1519-1530.
25. Banka et al. , Serum amyloid A (SAA): influence on HDL-mediated cellular cholesterol efflux. J Lipid Res. 1995;36(5):1058-1065.
Example 4
In this example, panel 18 (ApoA1, ApoC1, ApoC2, ApoC3, and ApoC4) was applied to samples from the Fairbanks Institute for Healthy Communities and a subset of the Abcodia Biobank.

The Fairbanks Institute for Healthy Communities (Fairbanks Institute) is creating an extensive annotated biospecimen repository for hypothesis-driven research. These biospecimens are linked with longitudinal clinical and disease-specific health record information extracted from the Indiana Network for Patient Care (INPC), a statewide electronic clinical data repository. The Fairbanks Institute CAD collection used includes sera and plasma from individuals diagnosed with coronary artery disease, as well as matched controls. Using this sample set, a predictive model of cholesterol withdrawal (Table 37B, Model 1) based on mass spectrometrically determined molar protein levels was established between healthy (control) patients (n=74) or coronary artery disease (CAD). Applied to a sample set consisting of previously diagnosed patients (n=157). Comparing the predicted cholesterol withdrawal values between each cohort, as shown in FIG. Whitney test). Running a logistic regression on the protein components of panel 18 to optimize the clinical outcome split (control vs. CAD) results in the parameters of Table 37, Model 3. We observe improved stratification of patient cohorts based on outcomes when applied to the Fairbanks clinical set. As shown in Figure 9, application of the outcome optimization model stratifies patients on a biomarker score measure that correlates with the risk of being diagnosed with CAD. Here, the CAD cohort shows significantly higher biomarker scores compared to the control group (P<0.0001 by Mann-Whitney test).

We also applied the model of panel 18 to a subset of the Abcodia Biobank. This subset consisted of 200,000 postmenopausal women aged 50-74, of whom 50,000 were longitudinally sampled over several years. In this study, control patients were matched to those with confirmed non-fatal myocardial infarction by sampling site, age, and sampling date. Both Model 1 and Model 3 (Table 37B) show a persistent and statistically significant divergence between these cohorts over a range of 0-8 years before the event.

All publications and patents mentioned in this application are herein incorporated by reference. Various modifications and variations of the described methods and compositions of the invention will be apparent to those skilled in the art without departing from the scope and spirit of the invention. Although the invention has been described in connection with specific preferred embodiments, it should be understood that the invention as claimed should not be unduly limited to such specific embodiments. . Indeed, various modifications of the described embodiments of the invention which are obvious to those skilled in the relevant fields are intended to be within the scope of the following claims.

Claims (15)

detectably labeled ApoA1 protein or A composition comprising a fragment thereof, wherein the method comprises:
a) detecting the level of one or more HDL-related proteins in a sample, wherein said one or more HDL-related proteins comprise ApoA1 protein or fragments thereof; and
b) generating a CVD risk score and optionally a CEC score using a first algorithm to determine an estimated risk of CVD in said subject, an estimated CEC of said sample, or a combination thereof; , the first algorithm comprises: i) multiplying each HDL-associated protein level by a predetermined factor to produce a multiplied value; and ii) summing the multiplied values to add a panel-specific constant value. performing operations thereby generating said CVD risk score and optionally said CEC score. Step b) further comprises generating a probability of CVD using a second algorithm, said second algorithm calculating said CVD risk score according to the following formula: CVD probability = 1/(1 + exp(- risk score )), and optionally further comprising multiplying the CVD probability by 100 to determine a CVD risk ratio. A CVD risk rate of less than 10% is considered low risk, a CVD risk rate of 10-20% is considered intermediate risk, and a CVD risk rate of greater than 20% is considered high risk. A composition for use as described in . A composition for use according to any preceding claim, further comprising c) generating a report indicative of said CVD risk score and optionally said CEC score. said one or more HDL-related proteins is apolipoprotein (Apo) C1, ApoC2, ApoC3, ApoC4, ApoD, ApoE, ApoL1, ApoM, CLU, HP, phospholipid transfer protein (PLTP), PON1, SAA1, and SAA2; A composition for use according to any one of claims 1 to 3, further comprising an HDL-related protein selected from the group consisting of or fragments thereof. Composition for use according to any one of claims 1-3, wherein said one or more HDL-associated proteins are proteins listed in biomarker panels 1-30. 2. A composition for use according to claim 1, wherein step a) comprises detecting the level of two or more HDL-associated proteins. 2. A composition for use according to claim 1, wherein step a) comprises adding a reagent to said sample, said reagent digesting said one or more HDL-associated proteins in said sample. 2. Composition for use according to claim 1, wherein step a) comprises mass spectrometry or immunoassay. Composition for use according to any one of claims 1 to 3, wherein the sample is selected from the group consisting of serum samples, plasma samples, blood samples and purified high-density lipoprotein samples. After step a) and before step b),
normalizing the level of the one or more HDL-related proteins with the estimated level of total HDL particles, or ApoA1, or HDL-cholesterol to produce a normalized value of at least one HDL-related protein. A composition for use according to any one of claims 1-3. said estimated level of total HDL particles, or ApoA1, or HDL cholesterol in a sample is determined by measuring the level of an internal standard added to the sample, said internal standard comprising labeled HDL-related proteins; A composition for use according to claim 11. 4. Any one of claims 1 to 3, wherein step a) comprises adding an internal standard to the sample and detecting said internal standard, said internal standard being a labeled or unlabeled HDL-related protein. A composition for use according to paragraph 1. 13. The composition for use according to claim 12, wherein said labeled HDL-associated protein is selected from the group consisting of Apo1, ApoC1, ApoC3, ApoD, ApoE, ApoL1, ApoM, PLTP, SAA1 and SAA2. 2. A composition for use according to claim 1, wherein said first algorithm comprises a logistic regression step.

Images